AmelieSchreiber commited on
Commit
90c60d5
1 Parent(s): 060e98e

Update README.md

Browse files
Files changed (1) hide show
  1. README.md +3 -5
README.md CHANGED
@@ -56,11 +56,9 @@ comprehensive. Using the notebook linked above should help further evaluate the
56
  This model is a finetuned version of the 35M parameter `esm2_t12_35M_UR50D` ([see here](https://huggingface.co/facebook/esm2_t12_35M_UR50D)
57
  and [here](https://huggingface.co/docs/transformers/model_doc/esm) for more details). The model was finetuned with LoRA for
58
  the binary token classification task of predicting binding sites (and active sites) of protein sequences based on sequence alone.
59
- The model may need more training, however it still achieves better performance on the test set in terms of loss, accuracy,
60
- precision, recall, F1 score, ROC_AUC, and Matthews Correlation Coefficient (MCC) compared to the models trained on the smaller
61
- dataset [found here](https://huggingface.co/datasets/AmelieSchreiber/binding_sites_random_split_by_family) of ~209K protein sequences. Note,
62
- this model has a high recall, meaning it is likely to detect binding sites, but it has a precision score that is somewhat lower than the SOTA
63
- structural models mentioned above, meaning the model may return some false positives as well.
64
 
65
  ## Overfitting Issues
66
 
 
56
  This model is a finetuned version of the 35M parameter `esm2_t12_35M_UR50D` ([see here](https://huggingface.co/facebook/esm2_t12_35M_UR50D)
57
  and [here](https://huggingface.co/docs/transformers/model_doc/esm) for more details). The model was finetuned with LoRA for
58
  the binary token classification task of predicting binding sites (and active sites) of protein sequences based on sequence alone.
59
+ Note, this model has a high recall, meaning it is likely to detect binding sites, but it has a precision score that is somewhat lower than the SOTA
60
+ structural models mentioned above, meaning the model may return some false positives as well. This may be undesirable for various reasons, one
61
+ of which could be cost of lab experiments where a higher precision might be more desirable.
 
 
62
 
63
  ## Overfitting Issues
64