README.md

metadata

tags:
  - generated_from_trainer
metrics:
  - precision
  - recall
  - f1
model-index:
  - name: prot_bert_bfd-disoDNA
    results: []

prot_bert_bfd-disoDNA

This model was trained from scratch on the None dataset. It achieves the following results on the evaluation set:

• Loss: 0.1323
• Precision: 0.9442
• Recall: 0.9717
• F1: 0.9578

Model description

This is a token classification model designed to predict the intrinsically disordered regions of amino acid sequences on the level of DNA disorder annotation.

Intended uses & limitations

This model works on amino acid sequences that are spaced between characters.

'0': No disorder

'1': Disordered

Example Inputs :

D E A Q F K E C Y D T C H K E C S D K G N G F T F C E M K C D T D C S V K D V K E K L E N Y K P K N

M A S E E L Q K D L E E V K V L L E K A T R K R V R D A L T A E K S K I E T E I K N K M Q Q K S Q K K A E L L D N E K P A A V V A P I T T G Y T D G I S Q I S L

M D V F M K G L S K A K E G V V A A A E K T K Q G V A E A A G K T K E G V L Y V G S K T K E G V V H G V A T V A E K T K E Q V T N V G G A V V T G V T A V A Q K T V E G A G S I A A A T G F V K K D Q L G K N E E G A P Q E G I L E D M P V D P D N E A Y E M P S E E G Y Q D Y E P E A

M E L V L K D A Q S A L T V S E T T F G R D F N E A L V H Q V V V A Y A A G A R Q G T R A Q K T R A E V T G S G K K P W R Q K G T G R A R S G S I K S P I W R S G G V T F A A R P Q D H S Q K V N K K M Y R G A L K S I L S E L V R Q D R L I V V E K F S V E A P K T K L L A Q K L K D M A L E D V L I I T G E L D E N L F L A A R N L H K V D V R D A T G I D P V S L I A F D K V V M T A D A V K Q V E E M L A

M S D K P D M A E I E K F D K S K L K K T E T Q E K N P L P S K E T I E Q E K Q A G E S

Training and evaluation data

Training and evaluation data were retrieved from https://www.csuligroup.com/DeepDISOBind/#Materials (Accessed March 2022).

Training procedure

Training hyperparameters

The following hyperparameters were used during training:

• learning_rate: 2e-05
• train_batch_size: 8
• eval_batch_size: 8
• seed: 42
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• num_epochs: 3

Training results

Training Loss	Epoch	Step	Validation Loss	Precision	Recall	F1
0.0213	1.0	61	0.1322	0.9442	0.9717	0.9578
0.0212	2.0	122	0.1322	0.9442	0.9717	0.9578
0.1295	3.0	183	0.1323	0.9442	0.9717	0.9578

Framework versions

• Transformers 4.21.3
• Pytorch 1.12.1+cu113
• Datasets 2.4.0
• Tokenizers 0.12.1