File size: 1,787 Bytes
a736b26
 
731175a
a736b26
 
 
 
 
9d40997
 
c508c1a
 
5e4aa2f
 
 
 
a736b26
c508c1a
52eadfe
c1bd2e4
52eadfe
c8fb47a
b46b3c2
 
 
 
52eadfe
c1bd2e4
52eadfe
27a9383
c1bd2e4
52eadfe
c1bd2e4
3f0dcb3
c1bd2e4
3f0dcb3
c1bd2e4
3f0dcb3
c1bd2e4
52eadfe
 
c1bd2e4
 
 
 
 
 
 
 
 
 
3c66691
c1bd2e4
efb0f11
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
---
base_model: westlake-repl/SaProt_35M_AF2
library_name: peft
---

# Model Card for Model ID

<!-- Provide a quick summary of what the model is/does. -->
This model is trained on a sigle site deep mutation scanning dataset and
can be used to predict fitness score of mutant amino acid sequence of protein [DLG4_RAT](https://www.uniprot.org/uniprotkb/P31016/entry) (Disks large homolog 4). 

Postsynaptic scaffolding protein that plays a critical role in synaptogenesis and synaptic plasticity by providing a platform for the postsynaptic clustering of crucial synaptic proteins.
Interacts with the cytoplasmic tail of NMDA receptor subunits and shaker-type potassium channels. 
Required for synaptic plasticity associated with NMDA receptor signaling. Overexpression or depletion of DLG4 changes the ratio of excitatory to inhibitory synapses 
in hippocampal neurons. May reduce the amplitude of ASIC3 acid-evoked currents by retaining the channel intracellularly.
May regulate the intracellular trafficking of ADR1B.


### Task type
protein level regression
### Dataset description
The dataset is from [Deep generative models of genetic variation capture the effects of mutations](https://www.nature.com/articles/s41592-018-0138-4).
And can also be found on [SaprotHub dataset](https://huggingface.co/datasets/SaProtHub/DMS_DLG4_RAT).

Label means fitness score of each mutant amino acid sequence.

### Model input type
Amino acid sequence
### Performance
 0.70 Spearman's ρ

### LoRA config
lora_dropout: 0.0

lora_alpha: 16

target_modules: ["query", "key", "value", "intermediate.dense", "output.dense"]

modules_to_save: ["classifier"]

### Training config
class: AdamW

betas: (0.9, 0.98)

weight_decay: 0.01

learning rate: 1e-4

epoch: 50

batch size: 128

precision: 16-mixed