id
stringlengths 24
24
| idx
int64 0
402
| paragraph
stringlengths 106
17.2k
|
|---|---|---|
67339b2df9980725cff94c52 | 34 | In this study, we developed the MultiModalTransformer (MMT) model, an innovative architecture for molecular structure elucidation that integrates multiple spectroscopic modalities including NMR, IR, and MS data. Our research demonstrates enhanced molecular structure prediction accuracy through the integration of diverse spectral data, achieving up to 94% correct identifications for experimental samples. We implemented a robust simulated data generation pipeline and an iterative improvement cycle, enabling effective performance on experimental data despite training only on simulations. The model's robustness is evidenced by its ability to overcome initial structural inaccuracies, maintaining substantial predictive power even with incorrect starting guesses. Additionally, the MMT provides explainable predictions through token-based analysis, offering insights into its decision-making process. While there's potential for further enhancement through increased experimental data training and automated peak-picking, the current model already serves as a powerful tool for chemists, effectively bridging the gap between simulated and experimental data. Its adaptability makes it particularly valuable for real-world applications where perfect initial structural information may not be available, representing a significant advancement in automated structure elucidation. |
63a03729ff46510499300ded | 0 | Drug discovery is a notoriously lengthy and costly process which is often described as finding a needle in a haystack. In this case, the haystack comprises on the order of 10 60 -10 100 synthetically feasible molecules , of which we need to find an ideal compound which satisfies multiple criteria including bioactivity, drug metabolism and pharmacokinetic (DMPK) profile, synthetic accessibility, etc. To reduce the number of compounds synthesized and improve efficiency during the drug development, Quantitative Structure Activity Relationship (QSAR) analysis has been applied for decades in the development of relationships between physicochemical properties of chemical substances and their biological activities to obtain a reliable mathematical or statistical model for prediction of the activities of new chemical entities. After Hansch proposed the QSAR concept , engineering molecular descriptors to build accurate models for the prediction of various properties has become the standard approach to QSAR modelling . Over the past few years, deep learning-based methods have achieved significant progress in molecular property prediction and de novo drug design. Recent results have shown that deep learning-based models can learn task specific molecular features directly from molecular structure and the conventional feature selection process is no longer needed. Among these efforts, Graphbased neural networks (GNNs) have particularly emerged as a powerful method for solving the property prediction problem. For example GAT , MPNN and recently attention based transformer models were shown to be able to provide more accurate prediction than traditional machine learning models. Among these models, usually atomic embedding representing atom features is learned from data and fused to form the graph embedding which is then correlated with endpoint properties. However, there is much less effort in exploring the atom pair embedding for building prediction model. Recently published UniMol model utilized atom pair embedding to build models, it learned atom pair representations based on 3D conformations and a resource demanding pre-training on millions of 3D conformation of unlabelled molecules was needed for achieving good performance on 2D based molecular property prediction tasks as well as 3D information based protein-ligand affinity prediction task. Here, we proposed a novel methodology called atom pair attention network (APAN) for learning atom pair embedding and applied it in multiple prediction tasks. One type of task was to build models on various benchmark datasets existing in MolecularNet database. These benchmark datasets contain purely 2D molecular information and have widely been used for developing novel deep learning algorithms. In contrast to the 3D based UniMol model, we built models using the atom pair embedding without doing pre-training on large number of unlabelled data and they out-performed previous state-of-the-art models. |
63a03729ff46510499300ded | 1 | The other type of task in current study was to build models for predicting proteinligand interaction based on complex structures. Accurate prediction of ligand binding affinity can help to prioritize compound synthesis and speed up the drug screening process . Typical scoring functions are divided into: physics-based methods , empirical-based methods and knowledge-based methods . They were used in popular docking software such as AutoDock Vina and GOLD etc. to rank 3D ligand poses. While the docking calculation can identify the binding pose of the protein-ligand complexes with relatively high accuracy, prediction of binding affinity and ranking compounds in the database are rather challenging . To improve the performance of these classical scoring functions, first wave of efforts focused primarily on applying Machine Learning (ML) method to develop scoring function. Ballester et al. introduced the RF-Score , which was the first scoring function relating the structural descriptors of complex with binding affinities using Random Forest regression model. The RF-Score achieved impressive scoring power comparing with classic scoring functions. Besides RF, other ML algorithms model were also proposed for training scoring function . One limitation of these machine learning scoring functions is their reliance on the feature engineering. Some recent efforts in developing descriptors for representing the chemical environ-ment of protein-ligand binding site were reported. Roitberg et al. modified the Behler and Parrinello symmetry functions to build single-atom atomic environment vectors (AEV), which were used as the input descriptors to construct a neural network potential named ANAKIN-ME (Accurate NeurAl networK engINe for Molecular Energies) . Barril et al. introduced Extended Connectivity Interaction Features (ECIF) to describe the types of protein-ligand atom pair and scoring functions using ECIF achieved outstanding scoring power. |
63a03729ff46510499300ded | 2 | Introducing deep learning models into scoring function development is a natural choice as the feature engineering step can be skipped . NNScore was the first reported deep learning-based scoring function, which was initially reported in 2010. It was simply constructed with feed-forward multi-layer perceptrons. Later, AtomNet , a docking tool built with 3D convolution neural network was reported. |
63a03729ff46510499300ded | 3 | Following on AtomNet, several works have attempted to apply 3D convolutional neural networks (3D CNNs) by extracting chemical information of each protein-ligand complex from 3D grids. However, these methods lack topological information of structures and are sensitive to the rotation of complex structures. In contrast to grid based methods, biomolecules can be treated as graphs and Graph Neural Networks (GNNs) can directly take graph as input. Unsurprisingly, various GNNbased scoring functions were reported, such as PotentialNet , graphDelta , SIGN etc. |
63a03729ff46510499300ded | 4 | As another application of our atom pair embedding method, we extended the atom pairing concept to 3D space by traversing protein-ligand atom pairs in binding site and utilized a special attention network, axial attention network, to learn atom pair embeddings in binding site and predict protein-ligand interaction. Our method outperformed some previous state-of-the-art models on the public dataset PDBBind 2016. |
63a03729ff46510499300ded | 5 | Our proposed APAN method adopts a general network architecture for learning 2D and 3D molecular representations, as shown in Fig. . Given a molecule M, its struc-ture can be represented either as a 2D graph M 2D or a 3D conformation M 3D . The model can take 2D molecular structure M 2D , ligand 3D conformation and proteinligand complex structure as input depending on if the task is to predict molecular property or protein-ligand binding affinity. The APAN contains two major layers, i.e. Atom Pairing Layer. For each molecule, its atomic feature is denoted as E ∈ R n * d , where n is the number of atoms, and d is the dimension of atom feature. We define several atom types representing the most common elements that can be found in organic compound: C, H, O, N, S, P, F, Cl, Br, I. Additionally, we also define important properties such as hybridization types, aromatic types, formal charge types and other atomic attributes of each atom. The atom feature E is defined as a one-hot vector representing those attributes. For atom i and j, their atom pair embedding is denoted as T ij and its initial value T 0 ij is set by the atom pair features E ij , which is obtained via an pairing function P (E i , E j ) operating on features E i (for atom i) and E j (for atom j). Detailed definition of pairing module will be introduced in the following sections. The pair representation and atom-pair communication were firstly proposed in the Evoformer in AlphaFold2 . Recently, the atom-to-pair communication was also proposed by Zhou et al. . Inspired by these works, we propose a pair-to-pair communication procedure and each pairwise embedding was updated by triangular consistency of pairwise embeddings, as shown in Fig. . Pair-to-Pair Communication. We establish the PPC process to learn molecular representation. Specifically, We use the pair embedding itself as the bias term in the multi-head self-attention mechanism to update atom pair embedding, as shown in Fig. . Given the atom pair T (l) ij as the input to the (l+1)-th block and define |
63a03729ff46510499300ded | 6 | Tij is taken as the latest atom pair ij embedding and goes to the next block as input. For protein-ligand complex structure, the atom pair embedding matrix is a nonsymmetric matrix and a specific axial attention block is used instead. The detailed explanation will be given in the following Section 2.5. |
63a03729ff46510499300ded | 7 | The atoms in the 2D molecular graph are explicitly connected by chemical bonds and we define M 2D = (V, E), where |V| = N refers to a set of N atoms and |E| = K refers to a set of K bonds in the M 2D . Firstly, we obtain the initial atom pair embedding T 0 ij through a pairing function P : |
63a03729ff46510499300ded | 8 | In order to incorporate bond information into the atom pair embedding, we design an atom-bond interaction network (ABIN) to improve the molecular representation by reinforcing message interactions between atoms and bonds in the graph, as shown in Fig. . Various MPNN methods have been proposed in learning molecular graph representation, and our ABIN module is inspired by the previous CMPNN (Communicative Message Passing Neural Network) method , which is a variant of MPNN architecture to improve molecular embedding by strengthening the message interactions between nodes and edges through a communicative kernel. |
63a03729ff46510499300ded | 9 | More specifically, the AGGREGATE module incorporates a message booster, which processes the bond infotmation with a max pooling function and calculates the element-wise product of the max pooling function's result and the sum of the hidden representations of bonds. The COMMUNICATE module is implemented as a multilayer perception. The INTERATION module adopts the Gated Recurrent Unit (GRU) |
63a03729ff46510499300ded | 10 | Here, E i and E j are the atomic features of the ligand atom pair (i, j). d ij is the Euclidean distance of atom pair (i, j). A(i, j, d ij ) is the affine transformation of the aggregated information of T ij and d ij , where W and B are learnable parameters. T ij is the D-channel atom pair (i, j) embedding. |
63a03729ff46510499300ded | 11 | The other scenario is to take protein-ligand complex structures as input to predict the protein-ligand binding affinity. Here, the core idea is to learn representations of key protein-ligand atom pairs for predicting protein-ligand interaction. Since a protein has large number of atoms, only atoms close to binding site are considered. In order to describe the protein-ligand binding site, we calculate the Euclidean distance between pocket atoms and ligand atoms. To reduce the computation cost, the pocket atoms are defined as protein atoms whose distance to ligand atoms d cutoff is within 3.5 Å. After preparing a set of pocket atoms, for each protein-ligand atom pair, as shown in Figure , a similar pairing process (from Equation 11-14) is employed to form initial atom pair embedding. The main difference to the formal scenario is that E i and E j are the atomic features of the pocket-ligand atom pair (i, j), instead of intra-molecular atoms. Another difference is that the ABIN blocks in the atom pairing layer are skipped, since protein pocket is not a single molecular entity. Once initial atom pair embeddings are obtained, they are passed into the PPC layer for making self-attention. One type of PPC module is built on the conventional Transformer encoder framework , where the self-attention unit is the main building block. For ligand specific tasks, the atom pair embedding tensor is symmetric and the normal self-attention block is used (as shown in Fig. ). For protein-ligand system, to reduce computation cost, we proposed a novel way to carry out self-attention. The atom pairs are arranged in a non-symmetric way, i.e. ligand atoms are listed in rowwise and pocket atoms are arranged in column-wise, as shown in Fig. (Left). To process such a non-symmetric information, recently proposed axial attention network is employed to learn the representation of ligand/protein 3D complex structures. |
63a03729ff46510499300ded | 12 | Axial attention includes column-wise and row-wise attention. More specifically, within the Axial attention module, the self-attention is first done along the vertical direction and then in the horizontal direction, corresponding to make attention along positions of protein atoms and ligand atoms individually. The specific network framework is shown in Fig. (Middle). Meanwhile, we also design a position encoding scheme in axial attention. Similar to the Graphormer model , we add row-wise attention related with 3D distance between pocket atoms and column-wise attention related with 3D distance between ligand atoms. The detailed model architecture is shown in Fig. (Right), where a row attention layer and column attention layer are arranged sequentially. The attention score A m ij denote the (i, j)-element of the Query-Key product matrix in the m-th row between atom pair h mi and h mj , we have: |
63a03729ff46510499300ded | 13 | In current study, three different APAN workflows were proposed for addressing various tasks. Fig. (a) shows the workflow using molecular 2D graph as input to predict ligand specific properties. Its atom pairing layer includes a concatenation block and multiple ABIN blocks. When it is desired to use ligand 3D conformations as input, the workflow in Fig. (b) can be utilized to make predictions on molecular properties. In this case, the atom pairing layer comprises an additional affine transformation block and a Gaussian kernel block to integrate the 3D distance into atom pair embedding. For 2D graph and ligand 3D conformation workflows, the normal Transformer self-attention block is employed in the PPC layer. Fig. (c) shows the workflow for predicting protein-ligand binding affinity, where the axial attention network block is employed in the PPC layer. Given the large dimension of atom pair embedding tensor, a pooling layer was adopted to reduce data dimension, In the pooling layer, two MEAN operations were done on the resulted atom pair embedding to obtain atompair features, where the average operation is performed on all the atom pair vectors. |
63a03729ff46510499300ded | 14 | Binding Affinity Prediction. The commonly used PDBBind 2016 was utilized as the dataset for binding affinity prediction. It contains three overlapping subsets: the general set, the refined set, and the core set. The 13,283 protein-ligand complexes in the general set were cleaned to form the refined set which contains 4,057 complexes of better quality, and the refined set was used for training. The known CASF-2016 set was chosen as the test set. In addition, we also selected CSAQ-HiQ as an additional test set . |
63a03729ff46510499300ded | 15 | Altogether eleven previously reported binding affinity models were selected for comparison, we make sure that the training set and division method used for these models are the same with ours. Here, the protein-ligand complexes in the refined set were split with a ratio of 9:1 according to the same seed of the SIGN model for training and validation. prediction tasks, QM7, QM8 and QM9 for Quantum Mechanical properties prediction tasks. Among these tasks, 3D ligand conformation could be chosen as the input data for the QM7, QM8 and QM9 datasets, while 2D molecular graph is the input for other seven datasets. |
63a03729ff46510499300ded | 16 | For a fair comparison, we utilized the same 5-fold cross-validation and division method for training all compared models in our own hands. So, for each task, we trained model five times and the prediction on test set was the average value from five models. The scaffold split method recommended by Yang et al. was employed to split the datasets into training, validation, and test, with a 8/1/1 ratio . The same criteria adoped by previous works were used to evaluate result: RMSE for regression tasks and ROC-AUC for classification tasks. |
63a03729ff46510499300ded | 17 | Implementation Details. Pytorch was used as the deep learning framework to implement our models and APEX 1 for the distributed training. We used the RAdam optimizer with learning rate 0.001 and weight decay for 0.011. Gradient was accumulated for 10 batches. Dropout was set to 0.1 and the early stopping steps was 30. Detailed hyper-parameters for model training can be found supporting materials. |
63a03729ff46510499300ded | 18 | • SIGN : A structure-aware interactive graph neural network, where the nodeedge aggregation process and interaction edges gathering process was implemented. It should be noted that the performance results for all the 11 reference models were directly taken from the SIGN paper, since the authors ran all these models in their own hands for benchmark purpose. Table shows the overall performance for all binding affinity prediction models including our APAN model. It can be seen that our approach achieved the best performance on both benchmark datasets among all the compared methods. Although our APAN model is only marginally better than the SIGN model which is the best model among all the baseline models, it clearly out-performs all other reference methods. |
63a03729ff46510499300ded | 19 | • CoMPT is a Communicative Message Passing Transformer (CoMPT) network which improves molecular graph representation by enhancing message interaction process between nodes and edges. Specifically, we compared the axial attention module with attention modules along single direction or totally skip the axial attention unit, while keeping the other parameters unchanged. Table shows the performance of the full model and models using individual attention component, and the model without using axial attention. It seems that the proposed axial attention module with dual attention blocks is clearly beneficial to the model performance. The dual attention mechanism can improve the performance by 14.7% in terms of RMSE and 11% in terms of correlation coeffi-cient to the row-attention model, 10.7% and 7.2% to the column-attention model, and 20.3% and 16.78% for the model not using axial attention at all. However, for ligand based tasks, the atom pair tensor is symmetric and there is no need to make axial attention, only conventional attention units are used. |
63a03729ff46510499300ded | 20 | The performance of our APAN models learning on 2D molecular graphs has been shown in Table . In MoleculeNet database, ligand 3D conformations are included in QM7, QM8 and QM9 datasets besides 2D molecule graphs. One interesting question would be if using 3D conformation will be beneficial for improving model performance. Previous work demonstrated that pre-trained 3D models performed very well for molecular property prediction. As illustrated in the 3D ligand workflow described in the method section, the Euclidean distance matrix between atom pairs can be incorporated in the atom pairing layer (as shown in Fig. ). Here, we would like to compare the performance between APAN model using 2D graph and the model using 3D information. |
63a03729ff46510499300ded | 21 | property data as prediction tasks and ligand 3D conformations are provided. The QM7 dataset contains molecules with up to seven heavy atoms, eight heavy atoms for the QM8 dataset, and nine for the QM9 dataset. These datasets provide the 3D Cartesian coordinates of the lowest energy conformations, as well as the geometric, energetic, electronic and thermodynamic properties which were all calculated by ab initio density functional theory (DFT) . QM7 contains one property, QM8 con-tains 12 quantum mechanical properties, and we selected homo, lumo and gap properties for the QM9 dataset following previous works . For each dataset, APAN models using 2D graph and using 3D information were built individually and their performance were compared. , N-Gram , PretrainGNN , GROVER , |
63a03729ff46510499300ded | 22 | GEM and Uni-Mol models, as well as our APAN models of 2D graph and 3D conformations. For QM8 and QM9 dataset, the MAEs are the average values over multiple tasks. As shown in As shown in Table , the APAN 3D models show better performance than the APAN models using 2D graph. As we know that those QM properties are closely related with ligand 3D conformation, it is understandable that models built on 3D conformation performs better than those with 2D graph for these tasks. Nevertheless, our APAN 2D models are still ranked as top model among all 2D related models which demonstrates that our APAN method works well for molecular representation learning. |
63a03729ff46510499300ded | 23 | Table : Ablation studies on models using coarse 3D conformations. Another interesting question was what will happen if ligand conformation is not necessary the relevant conformation to the predicted property? To answer this question, we took the compounds in the datasets of Table and generated their 3D conformation with the 3D conversion routine of RDkit and built APAN 3D models as we did in the QM tasks. It can be seen in Table that the performance of these 3D models is actually worse than the APAN 2D models. This highlights that using coarse/irrelevant 3D conformations didn't improve model accuracy comparing with 2D models, since they might bring in some noise to the model. This might be somehow contradict to the work on Uni-Mol model, but we should bear in mind that the Uni-Mol model was fine-tuned on a pre-trained 3D model which we didn't do here. |
63a03729ff46510499300ded | 24 | In current study, we introduced a novel attention method based on atom pair embedding and applied it on scenarios using 2D graph, ligand 3D conformation as well as protein-ligand complex as model input. Comprehensive benchmark studies on two types of prediction tasks, i.e. molecular property and protein-ligand interaction prediction, were carried out. Particularly, a novel axial attention unit was employed for predicting protein-ligand interaction. In general, our models achieved SOTA performance on all the endpoints among compared reference models. We also compared our APAN model using 2D and 3D ligand information. Our results show that using relevant ligand 3D conformation can improve model performance, while using irrelevant ligand 3D conformation in the model doesn't necessarily perform better than 2D based models. |
67619bfbfa469535b9e823c9 | 0 | Most artificial photosynthesis (AP) systems, including those at the forefront of research, focus on water splitting for hydrogen production or reducing CO₂ to basic hydrocarbons (C 1 -C 3 ). The ability to synthesize long-chain compounds (up to C 17 ) offers a transformative opportunity to produce higher-value chemicals, advancing AP technology scalability. |
67619bfbfa469535b9e823c9 | 1 | The mechanism involves forming photoactive Ti⁴⁺-OH bonds through the hydrolysis of Ti⁴⁺-Cl bonds by ambient air humidity. Upon exposure to light and air, these Ti⁴⁺-OH bonds release hydroxyl radicals, facilitating the reduction of the titanium center (Ti⁴⁺ → Ti³⁺/Ti²⁺). The reduced titanium species, with increased electron density, effectively capture atmospheric CO₂, initiating a spontaneous and intricate reduction process that produces long-chain oxygenated hydrocarbons. This solar-driven, autonomous production distinguishes the system from conventional AP methods. |
67619bfbfa469535b9e823c9 | 2 | Traditional AP approaches primarily target water splitting and CO₂ reduction to small hydrocarbons (C1-C3) using metal oxide catalysts. These methods often require high CO₂ concentrations and separate processes, achieving limited success. In contrast, the novel organotitanium-based system integrates water splitting and CO₂ reduction into a single, continuous process, producing oxygenated organics directly from ambient CO₂ (~0.04%). This capability to convert trace amounts of CO₂ into valuable materials represents a significant leap in AP technology. |
67619bfbfa469535b9e823c9 | 3 | A crucial feature of this system is the photocatalytic generation of hydroxyl radicals, released from Ti⁴⁺-OH, Ti³⁺-OH, and Ti²⁺-OH bonds. With a redox potential of 2.8 V (second only to fluorine), hydroxyl radicals are highly reactive, yet transient species that drive the organotitanium catalytic process. While TiO₂ photocatalysis is known to produce hydroxyl radicals, their synergistic role in facilitating complex reaction cascades in this system is unprecedented. |
67619bfbfa469535b9e823c9 | 4 | Although hydroxyl radicals typically exist on the nanosecond scale, they can persist for up to 600 seconds in aqueous environments, enabling diverse redox reactions. These include dimerization to form H₂O₂, oxidation of Ti=O bonds to peroxo-titanium species (Ti-O-O), oxidation of PI ligands to N-O functionalities, conversion of CO₂ to HCOOH, and condensation or radical coupling involving formaldehyde-derived intermediates. |
67619bfbfa469535b9e823c9 | 5 | The autonomous, multi-step nature of this system-driven by hydroxyl radical-mediated pathways-highlights its efficiency and versatility. Understanding these transformations is essential for optimizing or tailoring the system for specific goals, such as oxygen production via hydroxyl radical dimerization, where pH significantly influences outcomes. This system's ability to sustain multiple valuable reactions in a continuous process, especially DAC and CO₂ reduction, underscores its transformative potential. Further investigation into its mechanism could unlock even greater industrial and environmental applications. |
67619bfbfa469535b9e823c9 | 6 | Our research explores the synthesis of oxygenated hydrocarbon products with linear chains up to C 17 using an organotitanium photocatalytic system. This breakthrough enables the direct conversion of atmospheric carbon into complex, long-chain organic molecules, advancing green chemistry and renewable energy goals. The visible-light-activated organotitanium complexes exhibit self-organizing behavior, forming autocatalytic reaction networks. These networks drive three self-sustaining cycles, where reaction products catalyze subsequent reactions, showcasing an innovative approach to sustainable molecular assembly. |
67619bfbfa469535b9e823c9 | 7 | In the first photocatalytic cycle, the (PI)₂TiCl₄ complex serves as the essential "starter" catalyst, initiating a self-organized cascade. This complex interacts autonomously in the solid state with environmental inputs-visible light, H₂O, and CO₂-evolving into a sophisticated, multi-component photocatalytic system. Similar photocatalytic behavior has been observed in titanium click chemistry studies . |
67619bfbfa469535b9e823c9 | 8 | The process begins with hydrolysis of (PI)₂TiCl₄ by atmospheric moisture, forming intermediate complex 1, which converts to complex 2 (see Figure ). Under visible light, complex 1 reacts with CO₂, forming organotitanium carbonate complex 10 (illustrated in Figure ), while complex 2 releases hydroxyl radicals, generating Ti +2 complex 3. Complexes 1 and 3 capture atmospheric CO₂ (0.04%), a significant advance for climate change mitigation research. |
67619bfbfa469535b9e823c9 | 9 | Complex 4 coordinates with system-generated CO, replacing a PI ligand to form complex 8. This coordination facilitates CO insertion into the Ti-C bond, yielding the unique titanium α-ketocarboxylate complex 9. The four-membered Ti-O-C-C titanaoxetane ring structure, previously observed 9 in rhodium, silver, and other metals, appears for the first time in titanium chemistry as a ketocarboxylate ring. |
67619bfbfa469535b9e823c9 | 10 | The carbonate complex 10 (Figure ), formed from activated complex 1 and captured atmospheric CO₂, can generate formaldehyde (complex 12) and/or methanol (complex 13) through a hydrogenation mechanism. Hydrogen is produced internally via photocatalytic water splitting . The continuous catalytic cycle shown in Figure , incorporating DAC of CO₂, provides a sustainable source of formaldehyde and methanol as valuable feedstocks. |
67619bfbfa469535b9e823c9 | 11 | The THF-coordinated analog of complex 9, identified as complex 14 (Figure ) in MALDI-TOF analysis, initiates a new cycle in the cascade process, aiding the formation of long-chain organics. It uses formaldehyde as a building block, leading to the formation of complexes 15, 16, and 17. The THF coordination in complexes 14-17 is specific to the experimental conditions, as these were conducted in THF. In the absence of THF, complex 9 is expected to directly coordinate with formaldehyde. |
67619bfbfa469535b9e823c9 | 12 | The α-ketocarboxylate complex 14 undergoes a photocatalytic transformation to form ketene complex 15 and cumulene derivative 16 through a condensation with formaldehyde. The reaction involves dehydration via H₂O elimination, followed by hydrogenation to yield complex 17. While not detected by MALDI-TOF, it seems that a third condensation step occurs, likely contributing to the partially hydrogenated product. Complex 17, with four carbon atoms on the pendant chain, leads to the PI oxygenated analog (complex 21, Figure ), which is crucial in the third and final catalytic cycle. Titanium in complex 14 is highly electrophilic due to the electron-withdrawing chloride ligands and the carbonyl moieties of the titanaoxetane ring. This allows formaldehyde, with its nucleophilic oxygen, to react readily with the titanium center, forming complex 18 (Figure ). The second catalytic cycle begins with complex 14, which initiates a novel formaldehyde condensation process, driving the formation of long-chain organic molecules through a series of orchestrated steps outlined below: of such intermediate is supported by the detection of hydrogenated complex 17, characterized by a side chain of four carbon atoms, terminating in a reactive aldehyde group. |
67619bfbfa469535b9e823c9 | 13 | The structures of catalytic intermediates formed through the cascade transformations of the original (PI)₂TiCl₄ complex were primarily elucidated via MALDI-TOF MS (Matrix-Assisted Laser Desorption Ionization-Time of Flight Mass Spectrometry), which has proven highly effective in characterizing transition metal catalysts . MALDI-TOF is particularly advantageous for identifying intact metal complexes. Besides MALDI-TOF, the detailed process of deciphering this dynamic chemical system's cascade sequence and verifying individual complexes and other products was supported by other key observations as well: |
67619bfbfa469535b9e823c9 | 14 | (a) Comparative reactivity analysis of similar organotitanium systems, including hydroxyl radical formation by a substituted ansa-titanocene dihydroxido complex under daylight irradiation and the established ability of titanium oxo (Ti=O) groups to convert to peroxo species (Ti-O-O-) upon H₂O₂ exposure; (b) Observations of ordered reaction sequences across multiple sub-cycles of the photosynthetic cascade, as illustrated in Figures and (c) Expected formation of trichloro-and dichloro-titanium PI complexes upon daylight-induced reduction of the titanium tetrachloride PI coordination compound . |
67619bfbfa469535b9e823c9 | 15 | Complexes in Figure 1 , Figure and Figure are labeled with calculated molecular weights (MW) and expected m/z ratios, which reflect the natural isotopic distribution of the elements in each complex, along with the experimentally recorded MALDI-TOF MS data. The close match between calculated (MW, m/z) values and the experimental MS data strongly support the structural identity of each complex. |
67619bfbfa469535b9e823c9 | 16 | Figure displays the MALDI-TOF spectrum in CH₂Cl₂ for a (PI)₂TiCl₄ complex exposed to air and daylight over three weeks, highlighting mostly organic products. These products and their identities are discussed in detail. The relative concentrations of these materials are represented by the relative peak intensities in the MALDI-TOF spectrum, summarized in Table . The identification of each organotitanium catalytic intermediate was addressed in our previous publication 1 , which includes |
67619bfbfa469535b9e823c9 | 17 | The cascade operation continues with a third cycle, as shown in Figure . It begins with the N-oxidation of the PI ligand in complex 17, most likely facilitated by hydroxyl radicals, forming complex 21. This new sequence introduces a unique formaldehyde condensation process involving the aldehyde group on the side chain of the organotitanium complexes. This leads to the consecutive formation of complexes 22, 23, and 24, marking another step in the cascade transformation. A breakdown of the second sub-cycle highlights the following key steps: |
67619bfbfa469535b9e823c9 | 18 | The third cycle reaction sequence illustrates a novel formaldehyde condensation mechanism, enabling complex organotitanium structures to autonomously generate unsaturated α,ω-carboxylic acids-aldehydes. This process exhibits significant potential for additional carbon chain growth and radical coupling reactions. These advancements further underline the system's ability to construct intricate organic frameworks, offering transformative insights into organotitanium chemistry and catalytic cascade design, which will be elaborated upon in subsequent sections. |
67619bfbfa469535b9e823c9 | 19 | The hydrolytic separation of complexes 22, 23, and 24 yields three primary organic products, identified by MALDI-TOF as MS 128.1728, MS 140, and MS 152, respectively (Figure ). These products correspond to distinct stages of the chain-growth process. Furthermore, these intermediates demonstrate versatility through subsequent transformations, such as acid-catalyzed double-bond hydration, producing a hydrated derivative (MS 176.93), or partial hydrogenation, yielding MS 156. These pathways highlight the catalytic system's capability to generate a diverse array of organic molecules through targeted modifications of the initial products. The MALDI-TOF MS data points 199, 216, and 231.939 correspond to three sequentially hydrated C 9 unsaturated organic compounds, as illustrated in Figure . These structures are identified as hydrated derivatives of a parent C 9 unsaturated α-carboxylic acid-ω-aldehyde compound. However, the exact formation mechanism of this parent compound remains unclear and could not be conclusively determined. This ambiguity underscores a critical gap in understanding the cascade reaction process, necessitating further investigation into the pathways responsible for the generation of these hydrated organic derivatives.. The organic products corresponding to MS 216 and MS 231.939, both C 9 hydrated unsaturated derivatives, are involved in coupling reactions with a PI radical. In these reactions, the PI radical facilitates the extension or functionalization of the C 9 framework. |
67619bfbfa469535b9e823c9 | 20 | The next category of products includes C 12 to C 17 linear carbon chains, with a notably high concentration of C 16 organics. The conspicuous absence of C 10 to C 13 chains and the apparent upper limit at C 17 products are particularly striking. Interestingly, the C 12 products in this group originate from an unexpected radical dimerization of MS 128.1728, a C 6 compound mentioned earlier (Figure ). Hydrogen abstraction from the aldehyde functionality of this compound by hydroxyl radicals generates the corresponding C 6 radical, which subsequently dimerizes to form the C 12 α,ω-diacid (MS 253.909, Figure ). Similarly, the C 16 product (MS 306) arises from a partially hydrogenated C 8 monomer. In this case, aldehyde hydrogen abstraction by hydroxyl radicals produces a C 8 radical that dimerizes to yield an α,ω-dicarboxylic acid containing an unsaturated diketone functionality. These findings provide valuable insights into the radical-driven pathways that govern the formation of higher-order organic products in this system. A C 14 dimer originating from the C 7 MS 140 monomer could not be found, but a coupling product between a hydrogenated C 6 and C 7 monomer radical was identified yielding the C 13 dicarboxylic acid diketone MS 272, Figure . An unsaturated, partially hydrolyzed C 8 α,ω-hydroxylic acid keto radical, derived from complex 8, undergoes dimerization to form the C 16 unsaturated diacid-diketone-dialcohol (MS 342), as illustrated in Figure . |
67619bfbfa469535b9e823c9 | 21 | A third class of compounds produced by our photocatalytic system arises from radical-induced couplings between C 9 oxygenated keto radicals and PI radicals, yielding MS 407 and MS 423 (Figure ). The formation of these products aligns with the expected availability of PI radicals, as evidenced by the previously described radical oligomerization of PI during system operation 1 . |
67619bfbfa469535b9e823c9 | 22 | This class also includes two additional products, MS 457.947 and MS 474.12, which result from the coupling of related C 9 radicals with 5-position chlorinated PI radicals (Figure ). Notably, the coupling process appears to favor C 9 acid/alcohol unsaturated chain keto radicals over shorter counterparts (C 6 -C 8 ). This preference may be attributed to the longer half-life of C 9 radicals within the system. |
67619bfbfa469535b9e823c9 | 23 | Table organizes all products identified in the photocatalytic cycle, listed by their appearance in the MALDI-TOF MS spectrum (Figure ) along with their relative intensities, extracted using methylene chloride. These observations underscore the complexity and richness of the photocatalytic system in generating diverse organic products through a combination of radical coupling, oligomerization, hydrolysis, and hydrogenation processes. |
67619bfbfa469535b9e823c9 | 24 | The production of C 6 -C 17 oxygenated hydrocarbons from organotitanium compounds as described in this study represents a significant advancement in the field of organotitanium chemistry and artificial photosynthesis. Our work demonstrates that titanium-based catalysts, when properly tuned, can facilitate the formation of valuable, longer-chain oxygenated hydrocarbons under mild conditions. This method leverages the unique properties of titanium, particularly its ability to form stable intermediates while maintaining high reactivity, which has allowed for a cleaner and more direct pathway from simple starting materials, such as atmospheric CO 2 and water. |
67619bfbfa469535b9e823c9 | 25 | This research highlights the potential of organotitanium chemistry to bridge the gap between laboratory-scale artificial photosynthesis processes and practical applications in chemical synthesis and renewable energy. The capability to directly produce C 6 -C 17 hydrocarbons, which are critical intermediates in the chemical industry, positions this method as a promising alternative to conventional petrochemical routes. By generating these compounds from sustainable sources, we move one step closer to replacing fossil fuel-dependent methods with environmentally friendly and carbon-neutral processes. |
67619bfbfa469535b9e823c9 | 26 | Additionally, the ability to control the chain length and functionalization of these oxygenated hydrocarbons provides a versatile platform for synthesizing a range of chemicals, including biofuels, lubricants, and specialty chemicals. Future work will focus on further optimizing the catalyst system to enhance selectivity, yield, and overall efficiency. Moreover, improving our understanding of the reaction mechanisms at play will help in scaling up the process for industrial application, ensuring that it is not only sustainable but also economically viable. |
67619bfbfa469535b9e823c9 | 27 | In conclusion, this study opens new avenues for the use of organotitanium compounds in the generation of oxygenated hydrocarbons, contributing to the ongoing efforts to develop artificial photosynthesis technologies. The implications of this work extend beyond synthetic chemistry, offering a pathway to reduce carbon emissions and create more sustainable industrial processes. Future research will aim to address the challenges of scalability, economic feasibility, and environmental impact, with the ultimate goal of integrating this system into broader renewable energy frameworks. |
650f16cfb927619fe7aed0fc | 0 | One solution to this developing resistance problem is utilizing virulence inhibitors, which convert pathogenic organisms to benign ones by disarming them of their virulence. Many virulence inhibitors target external molecules or effectors, which can evade the development of antibacterial resistance caused by diminished permeability (where antibiotics become impermeable to the bacterial membrane as bacteria close protein channels) . In addition, it has been theorized that even when bacteria develop resistance to anti-virulents, it often results in nonfunctional virulent systems . Another benefit to virulence inhibitors is the relative lack of impact on the patient's microbiome: while traditional antibiotics often target both pathogenic and helpful bacteria, disarming the secretion system virtually exclusive to harmful gram-negative bacteria keeps intestinal flora intact and likely reduces gastrointestinal side effects . Salmonella, like many other gram-negative bacteria (Bordetella spp., Burkholderia spp., Chlamydia spp., Escherichia coli, Pseudomonas aeruginosa, Shigella spp., Vibrio cholerae, and Yersinia spp), utilizes the Type 3 Secretion System (T3SS) to secrete its bacterial effectors into the eukaryotic cell . The T3SS is a protein complex comprising over 20 different conserved proteins, creating a channel from the bacterial cell to the host cell , allowing it to "inject" bacteria effectors using a needle-like apparatus known as an injectosome . The T3SS contains 3 types of proteins: structural proteins, which form the body of the apparatus; translocators, which translocate the virulence factors into the host cells; and the effectors, which promote bacterial invasion and survival in the host cell . Its two continuous rings pass through the bacterial inner membrane, outer membrane, and peptidoglycan layer, anchoring the needle to the cell membrane. The structural apparatus contains the cytoplasmic complex, the export apparatus, the basal body, and the 2.5 nm needle (Figure ). To secrete the effectors, chaperone proteins form complexes with substrates, which are loaded onto free cytoplasmic complexes. SctK, SctQ, and SctL, specifically, are known as the "sorting platform" because of their involvement in recruiting chaperone-substrate complexes . The sorting platform then "unfolds" the effectors to "de-chaperone" them, and the ATPase in the cytoplasmic complex works with the export apparatus to guide and power the secretion of effector proteins through the needle . Effectors are passed through the translocon, which forms pores on the host cell membrane through the needle tip . This signals increased bacterial invasion and promotes bacterial survival, thus killing the host cell . |
650f16cfb927619fe7aed0fc | 1 | Figure : T3SS general structure with color-coded and labeled sections In Salmonella, the Salmonella Pathogenicity (SPI-1) virulence factor codes for the T3SS that enables bacterial invasion into host cells, where SPI-1 delivered effectors restructure the cytoskeleton of the host cell to force phagocytosis of the Salmonella pathogen . This process is critically dependent on the hexameric ATPase in the cytoplasmic complex, which is known as InvC in Salmonella. InvC is composed primarily of three folded domains: the N-terminal involved in the oligomerization of InvC 14 ; the conserved ATPase catalytic domain; and the C-terminal, which is potentially involved in recognizing chaperone-effector complexes . InvC plays a role in recognizing and binding the chaperone-effector complex, catalyzing ATP-fueled effector removal, and unfolding effectors in preparation for secretion . It is vital for Salmonella proliferation, for strains with loss-of-function mutant InvC were unable to successfully invade intestinal epithelial cells after the catalytic phosphate-binding loop motif (P-loop) was modified . |
650f16cfb927619fe7aed0fc | 2 | The necessity of InvC for virulence and its conserved structure across several pathogen species makes it an attractive target for T3SS inhibition. InvC is also highly related to several other T3SS ATPases, like Shigella Spa47, Yersinia YscN, E. coli EscN, and Pseudomonas HrcN, sharing 38-75% of its sequence with these orthologs -giving it potential as a target for broad-spectrum activity. The dysfunctionality of ATPase-knockout mutants is not restricted to Salmonella; in E. coli, deletion of the EscN gene (an ortholog to InvC) renders the pathogen unable to develop or secrete virulence factors . Similarly, in Y. enterocolitica, deletion of the YscN gene (another ortholog to InvC) abolishes the secretion of bacterial effectors . In Salmonella, when InvC is heavily destabilized, mutant strains are defective for translocation of effectors, bacterial egress, cytosolic colonization, and vacuolar replication . In addition, inhibition of T3SS ATPase avoids the risk of cross-reaction between T3SS ATPase inhibitors and human ATPase enzymes because bacterial ATPase and human ATPase share less than 25% of their sequences and their active sites are structurally different . Though there are many existing inhibitors of Salmonella T3SS, none have been shown to target the InvC protein vital for combating antibacterial resistance . |
650f16cfb927619fe7aed0fc | 3 | Computationally predicting the binding conformations of ligands to a receptor allows for the rapid screening of thousands of potential ligands to find the most viable lead compounds, an efficient precursor to traditional synthesis and biological assays . The compounds with the most negative binding energies (in kcal/mol) represent the most stable conformations; the lower the binding energy, the better a ligand can bind to the active site and is therefore predicted to be a more potent inhibitor. Here, we use these binding scores to identify lead compounds with the most potential. Additionally, catalogs of commercially available drug structures enable the targeted screening of compounds with known properties and metabolic activity. Especially with the availability of the Zinc database , we were able to preliminarily filter out small molecules with nonoptimal drug-likeness (an estimation of a compound's in-vivo potential based on its physiochemical properties) by selecting ligands with favorable molecular weight and lipophilicity-factors that have been demonstrated to impact molecular absorption, metabolism, and other factors (more details in methods/results) . |
650f16cfb927619fe7aed0fc | 4 | In this study, we screened thousands of potential ligands from the Zinc database and utilized AutoGrow4, an open-source structure-based drug design tool, to evolve potent inhibitors into novel drug-like compounds with optimized properties . We present over a dozen lead compounds derived from ten generations of ligand evolution with the highest binding affinities, as well as an evaluation of an entirely open-source route to novel inhibitor identification through the examination of the software's efficacy across generations. |
650f16cfb927619fe7aed0fc | 5 | We computationally screened selected molecules from the Zinc database against the active site of the InvC protein through AutoGrow4 for a total of 10 generations against the 6sdx model of InvC from the Protein Data Bank . Out of the Salmonella InvC models in the Protein Data Bank, we chose 6sdx because it contains ATP-gamma-s, the ligand used for investigation in the reference literature . For the binding site, we targeted the P-loop (amino acid residues 162-166) located in the ATPase catalytic core , as well as surrounding residues shown to interact with ligands in the binding site. When ATP-gamma S is bound to the enzyme, the phosphate groups form hydrogen bonds with G164 and T166, and a salt bridge with K165; the adenine group is stabilized by a pi-stacking bond with Y338 and forms a hydrogen bond with V411. As shown, this loop is highly relevant for ATP recognition and synthesis. |
650f16cfb927619fe7aed0fc | 6 | Since AutoGrow tends to create population homogeneity, convergence, and undesirable moieties after longer runs , we executed AutoGrow for 10 generations until improvement began to stall. We noticed compound fitness improving quickly in the first few generations and progress slowing in later generations, so compound suitability was unlikely to improve upon further experimentation. One limitation of this study is the slight drop-off in ligand fitness in the latter half of the generations, which reduced the versatility of our screened ligands. Additionally, we found that AutoGrow works best with well-researched proteins such as PARP-1 , so the relative novelty of 6sdx may also inhibit experimental efficacy. |
650f16cfb927619fe7aed0fc | 7 | Also shown is the structure of the amino acids in the binding site and their polar interactions with surrounding residues. The critical interactions between the binding site and the ligand in this pocket combined with the pocket's remodeling upon binding imply that inhibitors here would significantly impact InvC function . A shift from generation 0 to generation 3 is also present, though less obvious. Proportionally, there is a higher amount of ligands with binding affinities from -4.5 to -6.5 in generation 3 compared to generation 0. The peak in the generation 3 graph occurs at -5.5 to -6.0 kcal/mol, while the peak in the generation 0 graph occurs at -4.5 to -5.0 kcal/mol. From generation 3 to generation 7, the spread of the graph decreases, and the graph shows an increased proportion of compounds with binding affinities from -6.0 to -7.0. The peak of the graph remains at the same binding affinity. From generation 7 to generation 10, the proportion of compounds with binding affinities better than -5.0kcal/mol appears to decrease slightly, indicative of the previously discussed stagnation in later generations. The mean binding affinity of each generation was compared to the mean binding affinity of generation 0. We hypothesized that the mean binding affinity of generations 1-10 would be significantly smaller than the mean binding affinity of generation 0, setting a standard alpha-value of 0.05. The results of the significance test are shown below. mutations produced by AutoGrow4. The structure, highest binding affinity, and the first six digits of the compound code are provided in Figure . |
650f16cfb927619fe7aed0fc | 8 | While these compounds must be evaluated in vitro to reach definite conclusions, our novel ligands exhibit promising binding properties against Salmonella ATPase. Because heavy displacement of the P-loop is associated with ATP-analog binding, the disruption of those critical domains is likely to halt ATP hydrolysis . Without a functional ATPase enzyme, almost all pathogens are unable to properly inject effectors. This sabotages their ability to subvert host cell function, alter cytoskeleton structure, and evade immune system responses . Although the exact mechanisms linking ATP hydrolysis to the formation of the injectosome and secretion of protein effectors are not fully understood, several experiments have found a direct causal relationship between loss-of-function InvC mutations and loss of virulence . |
650f16cfb927619fe7aed0fc | 9 | In addition, the small molecules exhibit optimal molecular size and lipophilicity conditions that maximize clinical potential, making them promising agents against antibiotic-resistant Salmonella. Because many T3SS-dependent bacteria have similar secretion systems, the compounds identified here for Salmonella ATPase inhibition are favorable for broad-spectrum therapeutic development. The central T3SS ATPase among gram-negative bacteria is highly conserved, generating potential for T3SS inhibitors to treat a variety of infections. |
650f16cfb927619fe7aed0fc | 10 | The work presented also evaluates AutoGrow4's efficacy as a novel drug-design tool, for it is one of few open-source ligand evolution programs. AutoGrow4 enables discovery at a speed much faster than ex silico experimentation, with the capability to screen thousands of compounds and evolve them into a smaller group of energetically favorable ligands, introducing many possibilities for drug discovery and lead optimization. Although AutoGrow4 works less optimally without extensive knowledge of the target binding pocket and known inhibitors, there is still clear improvement in mean binding affinity across ten generations in our experiment. In addition, most compounds generated by AutoGrow are chemically feasible, and several filters (such as the Lipinski Strict Filter, the Ghose filter, or the VandeWaterbeemd Filter) can customize compound results . |
650f16cfb927619fe7aed0fc | 11 | Our study is limited by computational capabilities, available research on the ATPase InvC protein, and laboratory access. Thus, directions for future research include expanding the scope of generation 0 ligands and synthesizing the best-performing lead compounds for subsequent in-vitro evaluation. It is impossible to confirm a compound's inhibitory ability in pathogens only by knowing its computational binding efficacy. However, with protein structure modeling algorithms growing fast using Deep Learning and AI, AutoGrow4 should soon have even broader applications and provide more accurate results . |
650f16cfb927619fe7aed0fc | 12 | In order to obtain the center of the binding pocket and the size of the binding pocket, the open-source Python algorithm Scoria was used . Amino acid residues 162-167, 338, and 409-411 were used as the binding pocket, justified in the results section. This gave us a binding pocket center of (34.681, -13.6825, 10.3986) for x, y, and z respectively, and a binding pocket size with dimensions (12.0, 20.0, 12.0), for length, width, and height. |
650f16cfb927619fe7aed0fc | 13 | The initial pool of compounds is sourced from the ZINC-15 database, which has over 120,000,000 commercially available compounds for virtual screening. Selected compounds had a molecular weight between 150 to 250Da and a LogP value less than 5, with any level of reactivity. Compounds were then filtered with the Lipinski strict filter and converted into SMILE strings. The compounds were sorted into functional groups, and 100 compounds were randomly selected from each functional group. In total, this created an initial pool of 16,603 compounds (though generation 0 is slightly smaller due to AutoGrow's inability to dock certain compounds in time). |
650f16cfb927619fe7aed0fc | 14 | and ranks this new generation by docking each compound, taking the top molecules, or seeds, of each generation for the next. AutoGrow generates this new population through three different methods: an elitism operator, a mutation operator, and a crossover operator, which ideally creates a pool of compounds with better binding affinity than the previous. The 6sdx protein was processed and submitted as a .pdb, and the binding coordinates were obtained using Scoria, as explained previously. We ran AutoGrow for 10 generations, using MGLTools for conversion from .pdb to .pdbqt, and the default RDKit chemoinformatics reactions as the reaction library. |
650f16cfb927619fe7aed0fc | 15 | AutoGrow 4 uses QuickVina 2 as its default docking platform , which we did not modify. For molecular filters, we used the Lipinski Leniency Filter, as well as the Rank Selector for subsequent generations to prevent duplicates. In the first generation, we seeded 70 molecules from mutations and crossovers to the next generation alongside 100 elite molecules; for each subsequent generation, we seeded 50 molecules from mutations and crossovers, alongside 100 elite molecules for the next generation. The molecules from the Zinc database were first docked and scored to create a generation 0 before the selection process began for generation 1. This gave us a compiled folder of the molecules created each generation, their binding affinity, and the SMILE string for the molecule. The data was then compiled into graphs and tables for processing. |
674b5bbdf9980725cf9ca081 | 0 | The most relevant process of light-energy conversion in nature is photosynthesis, which takes place within the thylakoid membrane of cyanobacteria and chloroplasts. The thylakoid membrane is a lipid bilayer that assembles multiple protein supercomplexes in a transmembrane fashion to absorb light and transfer electrons across the membrane to generate high-energy products such as NADPH (nicotinamide adenine dinucleotide phosphate hydrogen) and carbohydrates as well as oxygen as a side product. The benefit of transmembrane electron transfer is that the oxidation site is physically separated from the reduction site and that such compartmentalization can fine-tune local reaction conditions to prevent detrimental recombination. Additionally, the transmembrane charge transfer generates a proton gradient which fuels the ATP synthase (ATP = adenosine triphosphate, see Figure ). Current artificial photosynthetic systems mostly investigate either an individual catalytic light-driven reduction (e.g., H2 evolution, 6 CO2 reduction ) or an individual oxidation reaction (e.g., water oxidation), both of which are still relying on the use of sacrificial reagents. A useful oxidation and reduction reaction combination is challenging due to their typically different reaction conditions and reported examples also rely on sacrificial reagents. For instance, reduction reactions such as H2 evolution perform best under acidic conditions due to a favorable thermodynamic redox potential connected to proton uptake. In contrast, oxidation reactions such as water oxidation are thermodynamically optimal at basic pH as they typically involve proton release. Therefore, it is desirable to perform the reduction and oxidation reaction in different compartments while connecting the compartments electronically. , under license number 5911880595918 from The American Association for the Advancement of Science), e) Molecular wire embedded in a 2-nm-thick SiO2 membrane enables electron flow from microbial catalysts to the inorganic anode. The system we report here uses a very simple, symmetric oligoaromatic molecular wire [1] 2+ that spans across the lipid bilayer membrane of artificial vesicles and effectively transfers electrons across the membrane under both inert and aerated conditions (Figure and). The design principle of [1] 2+ matches the one for membrane-spanning molecules, having a central hydrophobic region and two terminal hydrophilic groups. The distance between the terminal hydrophilic trimethylammonium groups is 3 nm and matches the thickness of the membrane to avoid destabilization of the membrane. Interaction with visible light is ensured via the benzothiadiazole at its center. Hydrophobicity is maintained by adding fluorene to its core. The rigid oligoaromatic design of the transmembrane molecular wire [1] 2+ has the potential for light-driven electron transfer in a more rigid polymer-based material, which will be relevant for the technological scale-up process. As electron transfer does not rely on diffusion within the membrane, it will be possible to transfer electrons in an environment like a polymer-based membrane. |
674b5bbdf9980725cf9ca081 | 1 | Pioneering work on light-driven transmembrane electron transport across vesicle bilayers was previously reported with a photosensitizer within the membrane combined with molecular electron shuttles resulting in diffusion-based transmembrane electron transfer. A prominent example of the pioneering work is shown in Figure . There are more diffusion-based systems, which are reported for inert conditions. Diffusion-independent light-driven electron transfer has been reported for photoelectrochemical studies with planar lipid bilayers using the type of photosensitizer shown in Figure . A few elegant, but synthetically demanding rigid transmembrane molecular wires and supramolecular barrel were reported for light-driven transmembrane electron transfer, shown in Figure , d. Apart from vesicle lipid bilayer membranes, the transmembrane design is also interesting for photoelectrochemical electron transfer and catalysis at electrode surfaces covered with an inorganic SiO2 membrane coating (Figure ). The function of these transmembrane electron transfer systems was to protect and connect the photo-and electrochemical components. |
674b5bbdf9980725cf9ca081 | 2 | The oligoaromatic light active molecule [1] 2+ was synthesized starting from benzothiadiazole, which was functionalized via bromination and successive two-fold Suzuki-cross-coupling with dimethyl fluorene boronic acid, yielding the unpolar core fbf (4,7-bis[9,9-dimethylfluoren2-yl]-2,1,3benzothiadiazole). Subsequent bromination, Suzuki-cross-coupling with a boronic acid of N,Ndimethyl aniline, followed by methylation of the nitrogen groups and ion exchange yielded the final product as PF6-salt as a yellow solid in overall 5 steps and an overall yield of 18 % (Figure ). The crystal structure can be found in the CCDC database with the identifier 2376492. The detailed synthesis and characterization can be found in the Supporting Information (Section S1). |
674b5bbdf9980725cf9ca081 | 3 | For liposome preparation, DPPC (dipalmitoylphosphatidylcholin) as the main lipid, [1] 2+ and the PEGylated lipid 14:0 PEG2000 PE were combined in the desired molar ratio of 100 :1 :1. DPPC liposomes were chosen because they are not permeable for larger water-soluble molecules at room temperature, which is the operating temperature during the experiments. In addition to DPPC, 1 mol% of 14:0 PEG2000 PE was added to the lipid mixture to increase stability and to prevent aggregation of the liposome via steric repulsion by the PEG chains on the liposome surface. Membrane integration of [1] 2+ into DPPC lipid bilayer was achieved via lipid film hydration, as described in detail in Section S2. Lipid film hydration yielded giant vesicles. When lipid film hydration was followed by extrusion, it yielded unilamellar vesicles with a typical diameter of 110 for empty liposomes and 140 nm average with encapsulated NADH according to dynamic light scattering (DLS). |
674b5bbdf9980725cf9ca081 | 4 | The size of the liposomes remained the same before and after irradiation for both cases, implying that the liposomes could retain their structure under irradiation (Section S3). To characterize the photophysical properties and membrane integration of [1] 2+ , UV-vis absorption and emission spectra were recorded, and emission lifetime measurements were performed (see Figure and Table ). It was observed that [1] 2+ has two absorption bands in polar organic solvent (acetonitrile) at 330 nm and 410 nm, respectively, and one emission band at 540 nm, which decays with a lifetime of τ = 5.4 ± 0.5 ns, indicating a spin-allowed fluorescence from a singlet excited state. |
674b5bbdf9980725cf9ca081 | 5 | In DPPC liposomes, the high-energy absorption band has the same shift as in the organic solvent. By contrast, the lower-energy band in the visible region shifts slightly bathochromically to 430 nm (Figure ), indicating a stabilization of the respective LUMO (lowest unoccupied molecular orbital) or destabilization of the respective HOMO (highest occupied molecular orbital), in the non-polar membrane environment, in line with similar dyes in unpolar solvents. The weak tail in the red region of the liposome absorption spectrum is due to light scattering and the Tyndall effect. In the emission spectra, a small hypsochromic shift towards higher energy of the emission band by 5 nm to 535 nm was observed, suggesting that the lowest excited state of [1] 2+ resides within the hydrophobic membrane core. The emission lifetime within the membrane remained similar with τ = 5.7 ± 0.6 ns; which is a low environmental influence and was also observed for other singlet emitters in solvents vs. membrane. Upon incorporation into the DPPC liposomes, which are less polar compared to acetonitrile, the absorption band at lower energies at around 410 nm undergoes a bathochromic shift to 430 nm, while the higher energy band remains at 330 nm as in acetonitrile. To rationalize why one of the bands shifts to lower energies while the other is barely affected, we resort to time-dependent density functional theory calculations. We used the CAM-B3LYP functional and def2-SVP set, as implemented in Gaussian 16, whereby the amount of HF exchange is reduced to reproduce the experimental absorption energies. Further details are presented in the SI Section S4. The calculations evidence that the two absorption bands should be assigned to the S1 and S3 singlet electronic excitations, respectively, while the transition to the S2 is dark. The hole and electron of the excitation to the S1 closely resemble the shape of the HOMO and LUMO (Figure ). During the excitation to the S1 (Figure ), electron density is transferred from the five six-membered rings closest to the center of the molecule to the most central moiety. By contrast, in the S3 (Figure ), both the hole and the excited state electron are similarly delocalized over large parts of the molecule, making the S3 comparable in electronic properties to the ground state. Further discussion of the electronic structures in the ground and excited states can be found in the SI Section S5. |
674b5bbdf9980725cf9ca081 | 6 | How these excited states are influenced by the environment requires elucidating the alignment of within the membrane. Furthermore, we also measure the integration of the hydrophobic chromophore core (fbf) in the DPPC membrane with confocal microscopy, which does not show a half-moon shape, presumably due to a random orientation of fbf within the lipophilic part of the membrane. Further exploration of the alignment of [1] 2+ in the membrane can be found in the SI section S6. The transmembrane alignment of [1] 2+ is ensured by its geometry matching with the geometry of the membrane. |
674b5bbdf9980725cf9ca081 | 7 | [1] 2+ has a hydrophobic core and two charged, hydrophilic N(CH3)3groups at a distance of 2.99 nm from one another based on crystal structure; in the structure optimized with DFT calculation the distance is 3.1 nm, and in the MD-simulations, the distance varies between 2.8-3.1 nm. As the membrane thickness is typically between 3-5 nm, the molecule can span across lipid bilayer as the thickness of the membrane, which is also demonstrated by MDsimulations, see a snapshot in Figure and other trajectories in the SI Section S7. The nature of this preferred orientation is unraveled with molecular dynamics (MD) simulations (see computational setup in Section S7). In those, the photoactive molecule in different ways, such as in a transmembrane fashion, or parallel to the surface of the membrane as we previously found for a different chromophore in a lipid bilayer. However, in our simulations Incidentally, this is exactly the region where both the excitations to S1 and S3 are located. As outlined above, the S1 exhibits a different electronic structure compared to the S0, and because the polar solvent will adjust to the ground state, it will stabilize the S0 more than an excited state of different electronic structure. When this stabilization is removed upon incorporation into the liposomes, the S0 will be more destabilized than the S1, resulting in the excitation redshift evidenced in the absorption spectra. The S3, by contrast, is similar to the S0 in electronic structure and will thus be similarly stabilized, explaining why it is much less affected by the environmental change and thus the excitation is not shifted noticeably. |
674b5bbdf9980725cf9ca081 | 8 | As a model reaction, we investigated the light-driven oxidation of NADH within the inner aqueous compartment of the liposome vesicles and the light-driven reduction of XTT (2,3-Bis(2-methoxy-4nitro-5-sulfophenyl)-2H-tetrazolium-5-carboxanilide sodium salt) in the outer aqueous bulk compartment. XTT is a well-known chemical reagent in biological assay to proof the presence of cellular reducing equivalents such as NADH. To enable transmembrane electron transfer across a lipid bilayer, we prepared DPPC liposomes that are impermeable to larger water-soluble molecules and integrated 1 mol% of the amphiphilic transmembrane chromophore [1] 2+ , as described above. The water-soluble electron donor NADH was added to the inner compartment of the liposome, while the water-soluble electron acceptor XTT was placed in the bulk or on the outer part of the liposome, forming so-called NADH/[1] 2+ /XTT liposomes (Figure ). Upon light irradiation, the absorption band at max = 470 nm gradually increased, corresponding to the electron uptake of XTT to generate its reduced form formazan (Fz), indicating that photoinduced electron transfer had occurred (Figure ). Simultaneously, the NADH absorption band at 340 nm is diminished, indicating oxidation of the electron donor. As there is significant spectral overlap with the absorption bands of [1] 2+ , the quantification of the transmembrane light-driven conversion was performed at the prominent Fz band at 470 nm. |
674b5bbdf9980725cf9ca081 | 9 | Leakage through the membrane was excluded through various reference experiments (Section S8), leading to the conclusion that, indeed, transmembrane electron transfer takes place. Based on the analysis of the thermodynamic properties, the excited state of [1] 2+ is reduced by NADH with a driving force of -0.53 V and a driving force of -0.98 V to oxidize by XTT. These values showed that it is easier to oxidize [1] 2+ than to reduce it (Section S11). |
674b5bbdf9980725cf9ca081 | 10 | It was furthermore confirmed that the original NADH/[1] 2+ /XTT liposomes perform light-driven transmembrane electron transfer under inert conditions with a rate constant of kAr = (1.76 ± 0.09) 10 -12 s -1 for Fz formation during the first 10 minutes of irradiation, and a quantum yield of 𝑄𝑦 = 0.045%. The rate constant of Fz was determined by fitting the first ten minutes of data with a linear function, and the 𝑄𝑦 was determined by the number of photons divided by the number of Fz formed as described in Section S9. Interestingly, we found that the presence of oxygen only reduced the reaction rate and 𝑄𝑦 by around factor 2 and the by a factor of 0.45 to 𝑄𝑦 = 0.020%, as reported in Table . This observation is very promising, as oxygen normally inhibits electron transfer reactions when performed with a diffusion-based electron shuttle. It seems that the transmembrane molecular wire approach provides robustness, potentially enabling the investigation of transmembrane water-splitting reactions in the future as well as reactions under aerobic conditions. |
674b5bbdf9980725cf9ca081 | 11 | Performing light-driven electron transfer under non-compartmentalized conditions, with electron donor and acceptor in the outer bulk ([1] 2+ /NADH, XTT in Figure and Table entry 2), significantly reduced the overall electron transfer efficiency, especially under aerated conditions. This suggests that the transmembrane electron transfer is indeed more efficient than a liposome-bulk system, and that compartmentalization might play a large role in accelerating the desired reactions, thereby minimizing the detrimental effect of oxygen. There are several reasons for such an improvement design is that a transmembrane potential can build up, which ultimately stops the transmembrane electron transfer. To balance the buildup of charges across the membrane, we tested two transmembrane ion transporters, gramicidin A and 18-crown-6 ether (also see Table , entries 3,4). |
674b5bbdf9980725cf9ca081 | 12 | Gramicidin A is an ion-channel-forming peptide, selective for small ions, such as protons. Under ambient aerated conditions, the initial quantum yield of gramicidin A was reduced and performed better in an inert atmosphere compared to the same system without gramicidin A. It was observed that gramicidin A played a role as an ion balancer, leading to an increase in initial quantum yield and an improvement in reaction rate. Gramicidin A accelerated the electron transfer rate constant by 11 % under Ar and 27 % under air. The quantum yield in the first 10 min is increased by 20 % under Ar and by 50 % under air. Another transmembrane ion transfer reagent 18-crown-6 ether was incorporated into the membrane. This transporter is selective only to K + cations. Adding 18-crown-6 to the phosphate buffer in the experiment enables the transport of K + cations to compensate for the charge imbalance in both compartments. It was observed that the addition of 18-crown-6 did not result in a better rate or quantum yield of transmembrane transfer compared to the addition of gramicidin A. |
674b5bbdf9980725cf9ca081 | 13 | To explore the photoinduced electron transfer in DPPC liposomes, the prepared samples of as a broad absorption band between 500 and 700 nm (see Figure ). Under oxidative conditions and upon photoirradiation, the formazan photoproduct Fz accumulates and obstructs the transient absorption measurements as it creates a bleach at the Fz absorption wavelength at 500 nm. The same artifact was also observed in the sample composition of NADH/[1] 2+ /XTT, making it challenging to evaluate the photo excited species in this composition. and XTT and also through the association of the anionic XTT with membrane-bound photosensitizer at the ground state (Figure ). As |
674b5bbdf9980725cf9ca081 | 14 | In summary, we report a very simple and novel design of a photoactive molecular wire for transmembrane electron transfer using visible light under inert and aerated conditions. We demonstrate, with confocal spectroscopy and molecular dynamic simulations, that [1] 2+ integrates with transmembrane orientation in the lipid bilayer. Light-driven electron transfer dynamics across the membrane were verified with two model redox reactions, and the excited state, photoreduced, and photooxidized products were characterized, along with the static and dynamic/static character of the initial photoinduced electron transfer. Overall, the liposomal system allowed us to electronically connect two complementary redox reactions through visible light-driven electron transfer. This process is crucial for solar energy conversion and the development of more advanced systems with robust and sturdy membranes. Additionally, it is important for the design of artificial cells and cellular compartments. While the challenge of resupplying the inner aqueous compartment with substrate still needs to be solved, the here presented findings on nano-reactors and transmembrane electron transfer are valuable for designing chemical synthetic methodologies involving compartmentalization. To create a larger photo-reactor, the same approach can be applied to polymer-based vesicles or larger planar membranes. The bio-related nature of the lipid bilayer membrane also offers opportunities for incorporating enzymatic catalysis, as the lipid membrane's complex units can be organized to carry out specific functions. |
674b5bbdf9980725cf9ca081 | 15 | To a 5 mL round bottom flask was added 1 mL of DPPC (5 mM in CHCl3), 1 mL of 14:0 PEG2000 PE (0.05 mM in CHCl3) and 1 mL of [1] 2+ (PF6)2 (0.05 mM in acetonitrile). The organic solvents were evaporated under vacuum leading to deposition of lipids on the flask wall. The film was dried under high vacuum for at least one hour and hydrated with phosphate buffer pH of 7.0 and added 15 mg of NADH. The dispersed lipid film was repeatedly freeze-thawing, using liquid N2 and a water bath at 52 °C, forming the giant vesicles. To obtain liposomes of uniform size, the dispersion was extruded at 52 °C through 200 nm cellulose membrane filters 11 times with an Avanti Polar Lipids miniextruder. The liposome mixture was subjected to a Sephadex G-25 size exclusion chromatography (SEC) column (6 cm length, 2 cm diameter) using phosphate buffer pH 7.0 as eluent; the liposome loaded with NADH was collected as the first band. |
674b5bbdf9980725cf9ca081 | 16 | l of 1 mM XTT, and 190 l phosphate buffer pH 7.0 were added. The mixtures were illuminated within a custom-made reactor equipped with four ventilators to exclude heating of the samples and an LED-stick (Nichia Corporation; NSPB500AS) (λ = 470 10 nm, 15.7 mW measured with a Hioki 3664 optical power meter set to 470 nm). The samples were monitored via UV-vis spectroscopy after 0, 1, 5, 10, 30, 60, 90, and 120 minutes. |
66016134e9ebbb4db97bb21e | 0 | Currently, the area of photovoltaics become revolutionized by newly emerging solar cells, organometal halide hybrid perovskites solar cell with its most recent efficiency jumps from 23% 1 into 29.15%. Remarkably, CH 3 NH 3 PbI 3 -based perovskites have very long electron and hole carrier diffusion lengths, high optical absorption, unique defect properties. Furthermore, recent theoretical and experimental results have shown that CH 3 NH 3 PbI 3 exhibits ambipolar self-doping properties: both good p-type and n-type conductivities can be realized by manipulating the growth conditions. Beside the exceptional high efficiency and V OC , lead halide perovskites are also capable of bandgap tuning from about 1.5 eV (CH 3 NH 3 PbI 3 ) to 2.2 eV (CH 3 NH 3 PbBr 3 ) by changing the compositions of perovskites. This capability makes lead halide perovskites as ideal candidates for all perovskites thin-film tandem cell and other optoelectronic device applications. |
66016134e9ebbb4db97bb21e | 1 | However, until now, it has been known that for all their gains in efficiency, organometal halide hybrid perovskites have faced stability challenges due to presence of defects, disorder and heterogeneity on photoexcited carrier dynamics. Mixed-cation perovskites solar cells have consistently outperformed their single-cation counterparts in both efficiency and stability cases. The first perovskites device to exceed 20% power conversion efficiency was fabricated with a mixture of methylammonium and formamidinium and improved stability. Recent reports have shown promising results with the introduction of cesium mixtures, enabling high efficiencies with improved photo-, moisture and thermal stability. McGehee 12 reported exceptional stability for new perovskites recipes that replace an organic component called methylammonium with formamidinium or the element cesium. The combination of the more stable perovskite and the protection provided by the ITO layer allowed the cells to operate for more than 1000 hours in air and 40% humidity without significant degradation. When encapsulated to protect them from moisture, these cells showed no sign of degradation for six weeks, even when exposed to temperatures of 85°C and a relative humidity of 85% for 1000 hours, a standard test of durability. -Panels that pass it usually will not fail due to heat and humidity over 25 years outside,‖ McGehee says. The increased moisture and thermal stability are especially important for various reasons including deposition of metal oxide contacts through atomic layer deposition or chemical vapour deposition that may require elevated temperatures or water as a counter reagent. With all these concerns, it is reasonable to find an alternative perovskites material that can improve the photo-, moisture and thermal stability issues of methylammonium based organolead halide hybrid perovskites solar cells. This is due to the reasons that methylammonium cation is weakly stable towards moisture and thermal treatments and it is more preferable to find materials with reasonable stability and relatively high efficiency for practical applications. The materials community is currently witnessing a fundamental change in the way novel materials are designed and discovered. A steady increase in computational power, accompanied by developments in quantum theory and algorithmic breakthroughs that allow for efficient yet accurate quantum mechanical computations, opens the door to computing properties of a wide range of materials that once seemed prohibitively expensive. As a result, explorations of the vast chemical space are increasingly being pursued and have significantly aided our intuition and knowledge-base of material properties. In our previous work we synthesized bulk crystals of MAPbI 3 . We, herein, investigated crystal structure, electronic and optical properties of CuPbI 3 , CuPbBr 3 , CuPbCl 3 and AgPbI 3 fully inorganic new perovskites for the first time, with the purpose for more stable and optimal light absorption compared to the methylammonium based organometal halide perovskites. First Principles calculations were performed based on DFT calculation as implemented in the VASP program during which the spin-orbit effect was neglected. These materials are direct bandgap material, which allow direct band to band transition with theoretically calculated tunable bandgap ranged 1.54 eV-2.33 eV (theoretically calculated) and 1.42-1.93 eV(experimentally determined). This suggests that these materials are suitable for photovoltaic and other optoelectronic device applications such as laser, light emitting diode and photovoltaic devices, which require direct band to band transitions. |
66016134e9ebbb4db97bb21e | 2 | Crystal structures/Geometric properties: Previous reports indicated that ABX 3 type compounds show various phases under different temperatures, but in the high temperature phase, all adopt a cubic perovskites structure , with a three-dimensional framework of corner-sharing BX 6 octahedron. Figure shows the crystal structure for cubic CuPbI 3 perovskites. The primitive cell of this perovskites is consisting of cornerlinked PbI 6 octahedral of the iodine atoms, with the Pb cation at their center and the Cu cation between them. The structure of CuPbI 3 was relaxed first and the obtained lattice constant for this cubic material is 6.405Å after optimization with α = γ = β = 90.0° for the unit cell shown in Figure . Volume of the cell is 6.405 ^3 Å ^3. The bond lengths obtained from the relaxed structure are: Pb -I: 3.21, Pb -Pb: 6.42, Pb -Cu: 5.55, Cu -I: 4.53, Cu -Cu: 6.41, I -I: 4.53 Å as well as the effective coordination number is 6.00. The obtained space group was also determined to be Pm-3m. The calculated crystal structure parameters are presented in Table Summary of crystal structure parameters The tolerance factor is invented by Goldschmidt to describe the perovskites structure by taking three types of spheres in a perovskites-structure into account. Since the ionic radii are generally known, a tolerance factor can be calculated, that indicates the degree of compatibility of a given set of ions with the ideal, cubic perovskites structure. This relationship was first explored by Goldschmidt, in 1926 , who suggested that it could be used to predict the likelihood that a pair of ions would form a perovskites structure phase. |
66016134e9ebbb4db97bb21e | 3 | The t value >1 implies the A cation is too large to fit in to their interstices and possible structures are hexagonal polytype; t = 0.9-1.0 indicates ideal condition with cubic perovskites; 0.71-0.9 -A cations are too small to fit in to their interstices with several possible structures such as orthorhombic and rhombohedral variants and t< 0.71 indicates A and B ions have similar ionic radii with possible close packed structures. For this purpose, it is assumed that for a stable structure to form the cations, just touch the surrounding anions (Goldschmidt's rule), then: t = (r A + r X )/(r B + r X ) = √2 or, where t is called the tolerance factor, r A is the radius of the cage site cation, r B is the radius of the octahedrally coordinated cation and r X is the radius of the anion. Goldschmidt's proposal was that a stable perovskites structure phase would form if the value of the tolerance factor, t, was close to 1.0. However, this is applicable at room temperature to the empirical ionic radii. The tolerance factor for CuPbI 3 at room temperature is 0.801 nm which could be tetragonal structure. The cubic unit cell edge, a, is equal to twice the B─X bond length: 2(B-X) = a. The width of the cuboctahedral cage site, √2a, is equal to twice the A─X bond length: 2(A-X) = √2a. This means that the ideal structure forms when the ratio of the bond lengths is given by: (A-X)/(B-X) = √2. .Because many perovskites structures have been described, it is now usual to use the measured bond lengths in the crystal rather than ionic radii to give an observed tolerance factor t obs : t obs = (A-X)/√2(B-X), where (A─X) is the average of the measured bond lengths between the A cation and the surrounding 12 anions and (B─X) is the average of the measured bond lengths between the B cation and the surrounding six anions. For CuPbI 3 perovskites, a is 2x3.2025 = 6.405Å, width of the cuboctahedral cage site is 2x4.53 equal to 9.06 Å and the tolerance factor given by the average bond length is 1.00 for all the perovskites materials. The stable structure is cubic structure for this material at high temperature. Error! Reference source not found. shows cubic crystal structures of CuPbX 3 derived by replacing . We find that the lattice constants of ABX 3 increase with increasing the size of X from Cl, to Br and I. respectively. This optical band gap value also agrees with the band gap value calculated for phase change in CH 3 NH 3 PbI 3 by UV-Vis spectroscopy. It is also reported that the presence of methylammonium cation or Cs + does not affect the band gap region and it is the inorganic component of lead halide perovskites that play a dominant role in ascertaining the band gap energy. Similarly, it has been reported that replacement of CH 3 NH 3 + or Cs + by a univalent transition d 10 metal ions, e.g., Cu + did not affect the band gap. Naeem et.al. revealed that the role of Cu + with d 10 configuration is just to balance the charge (Cu acts just as a charge donor here) and does not have any important contribution for the conduction and valence band states except for donating one electron to the Pb-I framework as is reported for CH 3 NH 3 + and Cs + ions in CH 3 NH 3 PbI 3 and CsPbI 3 , respectively. However, our result is opposite to their conclusion that Cu is predictable to directly influence the electronic properties, e.g. Cu + has contributions to the band edges of the CuPbX 3 materials. Figure shows the effect of size on the optical bandgap of ternary lead halide perovskites. We observed that the band gap increase with increasing size of A ions and decreasing size of X ions. These chemical trends are important to understand and optimize this type of solar cell absorbers materials. We explain these trends by the above band structure analysis shown in Figure , Figure The valence and conduction bands of methylammonium cation based perovskites are determined by the PbI 6 inorganic octahedron. The organic cation does not contribute to these bands but modulates the band gap by modifying the lead-halide bond distance; as a result, the shifts produced by substituting the organic cation are less pronounced than those produced by substitution of the halide. The full tunability of the band gap for CH 3 NH 3 PbX 3 has been demonstrated for I-Br compositions and strongly suggested for Br-Cl compositions. The valence and conduction bands of Cu + and Ag + based ternary perovskites are not only determined by the inorganic PbI 6 octahedron, but also by the Cu + and Ag + . The role of Cu + and Ag + is not only to modulate the band gap but also by tuning the band structure. Therefore, we demonstrate the full tunability of CuPbX |
676d46fd6dde43c9086402b6 | 0 | Hydrogels have emerged as pivotal materials in biomedical applications, particularly in the realms of bioprinting and tissue engineering. These hydrophilic polymer networks can retain substantial amounts of water while maintaining structural integrity, closely mimicking natural tissues' extracellular matrix (ECM). Their biocompatibility, high water content, and capacity to encapsulate cells and bioactive agents make them ideal candidates for creating biomimetic tissue constructs. However, the diverse mechanical and functional requirements of various tissues necessitate the development of hydrogels with tunable properties. Gelatin, a denatured form of collagen, is widely used in hydrogel formulations due to its biocompatibility and cell adhesion properties. Methacrylation of gelatin (MG), also widely known as GelMA , makes it photo-crosslinkable, enabling stable hydrogel formation under UV light in the presence of a photoinitiator. This characteristic makes methacrylated gelatin (GelMA) particularly suitable for advanced bio-fabrication techniques like digital light processing (DLP) and extrusion-based 3D printing, where cross-linking control is crucial. Yet, single-component hydrogels often lack the mechanical robustness required for loadbearing applications or long-term stability. To address this limitation, the incorporation of microparticles into hydrogels has emerged as a promising strategy for enhancing these properties without compromising biocompatibility. In this study, we introduce a composite hydrogel system that incorporates precrosslinked porcine gelatin microparticles within a methacrylated fish gelatin (MFG) matrix. This hybrid approach was specifically chosen to address the physical property differences between porcine and fish gelatins, with direct implications for 3D printing performance. Notably, porcine skin gelatin possesses a higher gel strength and gelation temperature than fish gelatin, causing it to solidify rapidly at room temperature. Such behavior presents significant challenges for 3D printing, which requires materials to be in a fluid or semi-fluid state to enable precise deposition and shaping. If porcine skin gelatin were directly mixed with fish gelatin, the composite would rapidly increase in viscosity and could even solidify at room temperature, resulting in a material unsuitable for DLP and extrusion printing. By pre-crosslinking the porcine gelatin into microparticles and subsequently dispersing them in a methacrylated fish gelatin matrix, we achieve a composite hydrogel that retains fluidity until UV cross-linking. This structure provides essential shear-thinning behavior and stability suitable for various 3D printing modalities, including embedded printing, a technique in which bioinks are extruded into a support bath to construct intricate tissue structures. Embedded printing has gained considerable traction as a method for achieving detailed, multilayered designs that require additional support during fabrication. Our microparticlereinforced hydrogels demonstrate the essential characteristics for this application, with the mechanical and rheological properties necessary to act as support baths. |
676d46fd6dde43c9086402b6 | 1 | Gelatin's ability to be chemically modified makes it an attractive material for tissue engineering. Crosslinking gelatin microparticles enhance their stability, allowing them to resist rapid degradation while maintaining their functional properties within the hydrogel matrix. . Crosslinked gelatin microparticles can also influence other important hydrogel properties, such as swelling behavior and degradation rate. By tuning the size and crosslinking density of the gelatin microparticles, it is possible to control the water uptake and swelling behavior of the hydrogel composite. This is particularly important in applications where precise control over hydrogel swelling and degradation is required, such as drug delivery systems or tissue scaffolds that must degrade at a specific rate. This study aims to demonstrate the potential of composite hydrogels reinforced with crosslinked microparticles as advanced materials for bio-fabrication, with tunable mechanical and functional properties that extend their applicability beyond conventional single-component hydrogels. To confirm the structural integration and performance of the composite hydrogels, various characterization techniques have been employed. Nuclear Magnetic Resonance (NMR) spectroscopy was used to verify the degree of methacrylation of both gelatins, a key step in ensuring the material's photo-crosslinkable properties. Scanning Electron Microscopy (SEM) provided detailed images of the pore structure and porosity within the hydrogels, critical for evaluating the microstructural arrangement influenced by the microparticles. Mastersizer analysis was used to assess the size distribution of the microparticles, ensuring consistent particle dispersion throughout the hydrogel matrix. |
676d46fd6dde43c9086402b6 | 2 | Further analysis included mass swelling measurements to evaluate the hydrogels' water uptake capacity and mechanical testing to assess compressive strength and stiffness. Thermal properties were characterized by Thermogravimetric Analysis (TGA), Differential Thermal Gravimetry (DTG), and Differential Scanning Calorimetry (DSC) to investigate the thermal stability of the composites, essential for applications requiring durability under variable conditions. Rheological assessments were conducted to understand the shear-thinning behavior crucial for extrusion-based 3D printing, while UV-Vis spectrometry confirmed effective photocrosslinking at wavelengths compatible with Digital Light Processing (DLP) printing. |
676d46fd6dde43c9086402b6 | 3 | The DLP technique enables precise 3D structuring of hydrogel objects by selective photocrosslinking the methacrylated matrix, showcasing the potential of these hydrogels in fabricating complex tissue engineering constructs and other biomedical devices. Additionally, the microparticle-reinforced hydrogel's shear-thinning and stability properties make it wellsuited as a support bath in Embedded 3D printing, where bioinks can be extruded into the support bath to form complex, multilayered constructs with high spatial fidelity. |
676d46fd6dde43c9086402b6 | 4 | The integration of crosslinked gelatin microparticles into methacrylated gelatin hydrogels offers a promising avenue for developing advanced composite hydrogels with tailored mechanical, thermal, and rheological properties. These hybrid systems provide the flexibility to fine-tune the hydrogel's performance characteristics, making them ideal for bio- fabrication applications where structural integrity, mechanical resilience, and biocompatibility are critical. |
676d46fd6dde43c9086402b6 | 5 | The detailed experimental procedure has been described previously. In short, type A gelatin derived from porcine skin tissue was dissolved in CB (Carbonate-Bicarbonate) buffer (0.1 M buffer comprising 3.18 g sodium carbonate and 5.86 g sodium bicarbonate in 1 L distilled water), and the pH was adjusted with 5 M sodium hydroxide or 6 M hydrochloric acid. Subsequently, MAA (Methacrylic anhydride) was added to the gelatin solution under magnetic stirring at 500 rpm. The reaction proceeded for 1 h, and then the pH was readjusted to 7.4 to stop the reaction. After being filtered, dialyzed using Spectra Por S/P 2 Dialysis membrane (12 -14 kDa, 25 mm flat width) (Cole-Parmer, USA), and lyophilized, the samples were stored at -20°C until further use. The standard conditions of the synthesis were: CB buffer at 0.25 M, initial pH adjustment at pH 9, MAA amount at 0.01-0.1 mL per gram of gelatin concentration at 10 w/v%, reaction temperature at 50°C, and reaction time for 1 h. |
676d46fd6dde43c9086402b6 | 6 | A solution of 1% (w/v) 2-hydroxy -1-(4(hydroxyethoxy)phenyl) -2-methyl -1propanone (Irgacure2959) photo-initiator was dissolved in MilliQ water at 80°C and stirred until fully dissolved. The freeze-dried functional gelatin macromer was weighed and dissolved in the above solution containing a photo-initiator to prepare prepolymer solutions with the desired protein concentration of 10% (w/v) under stirring at 40°C. |
676d46fd6dde43c9086402b6 | 7 | To prepare microparticles, 65g of prepolymer solution was poured into an aluminum dish and was photo-polymerized as described above. The crosslinked hydrogel was suspended in 30 mL of Milli-Q water and was poured into a commercial laboratory blender (Waring, USA) for chopping into finer particles. The slurry was blended at 15,000 rpm speed for 5 min. |
676d46fd6dde43c9086402b6 | 8 | Blending at such high speeds created a layer of foam, as air bubbles were introduced. The blended slurry was vacuum filtered through a 70 μm (nylon) cell strainer (VWR, USA). The filtered slurry solution was then poured into disposable conical-bottom 50 mL centrifuge tubes (VWR, USA). The tubes were centrifuged at 10,000 rpm at 4°C for 10 min in an Avanti J-26 XPI centrifuge (Beckman Coulter, USA). The slurry was washed 3 times with MilliQ water, and a pellet of fine granules was collected at the bottom of the tube when the water was aspirated. |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.