README.md

---
base_model: westlake-repl/SaProt_35M_AF2
library_name: peft
---
# Base model: [westlake-repl/SaProt_35M_AF2](https://huggingface.co/westlake-repl/SaProt_35M_AF2)

# 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 AsCas12f. 

## Protein Function
AsCas12a is widely utilized as genome-editing tools in human cells.


### Task type
protein level regression

### Dataset description
The dataset is from [An AsCas12f-based compact genome-editing tool derived by deep mutational scanning and structural analysis](https://doi.org/10.1016/j.cell.2023.08.031).


Label means fitness score of each mutant amino acid sequence. 
Ranging from negative infinity to positive infinity. The wildtype sequence has fitness 1. 
If the effect larger than 1 represents high fitness, smaller than 1 represents low fitness.

### Model input type
Amino acid sequence
### Performance
 0.60 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: 36

precision: 16-mixed