- ProteinLM
- Overview
- Guidance
- Project Structure
- Usage
- Downstream Tasks Performance
- Citation
- Contact
- Reference
We pretrain protein language model based on Megatron-LM framework, and then evaluate the pretrained model results on TAPE (Tasks Assessing Protein Embeddings), which contains a set of five biologically relevant semi-supervised learning tasks. And our pretrained model achieved good performance on these tasks.
The proposal of pre-training models such as Bert have greatly promoted the development of natural language processing, improving the performance of language models. Inspired by the similarity of amino acid sequence and text sequence, we consider applying the method of pre-training language model to biological data.
We provide pretrain and finetune code in two separate folders. If you use the pretrained model we provide, you can simply download the checkpoint and follow the finetune guide. If you want to pretrain your own model yourself, you can refer to the pretrain guide.
For the pretrained model with 200 million parameters, you can download model checkpoint via GoogleDrive, or TsinghuaCloud.
For the pretrained model with 3 billion parameters, you can download model checkpoint from here.
.
├── pretrain (protein language model pretrain)
│ ├── megatron (model folder)
│ ├── pretrain_tools (multi-node pretrain)
│ ├── protein_tools (data preprocess shells)
└── tape
├── conda_env (conda env in yaml format)
├── converter (converter script and model config files)
├── scripts (model generator, finetune)
└── tape (tape model)
As the structure above shows, there are two stages as follows.
- Pretrain
- Prepare dataset (
PFAM) - Preprocess data
- Pretrain
- Prepare dataset (
- Finetune
- Convert pretrain protein model checkpoint
- Finetune on downstream tasks
Detailed explanations are given in each folder's readme.
| Task | Metric | TAPE | ProteinLM (200M) | ProteinLM (3B) |
|---|---|---|---|---|
| contact prediction | P@L/5 | 0.36 | 0.52 | 0.75 |
| remote homology | Top 1 Accuracy | 0.21 | 0.26 | 0.30 |
| secondary structure | Accuracy (3-class) | 0.73 | 0.75 | 0.79 |
| fluorescence | Spearman's rho | 0.68 | 0.68 | 0.68 |
| stability | Spearman's rho | 0.73 | 0.77 | 0.79 |
Please cite our paper if you find our work useful for your research. Our paper is can be accessed here.
@article{DBLP:journals/corr/abs-2108-07435,
author = {Yijia Xiao and
Jiezhong Qiu and
Ziang Li and
Chang{-}Yu Hsieh and
Jie Tang},
title = {Modeling Protein Using Large-scale Pretrain Language Model},
journal = {CoRR},
volume = {abs/2108.07435},
year = {2021},
url = {https://arxiv.org/abs/2108.07435},
eprinttype = {arXiv},
eprint = {2108.07435},
timestamp = {Fri, 20 Aug 2021 13:55:54 +0200},
biburl = {https://dblp.org/rec/journals/corr/abs-2108-07435.bib},
bibsource = {dblp computer science bibliography, https://dblp.org}
}
If you have any problem using ProteinLM, feel free to contact via mr.yijia.xiao@gmail.com.
Our work is based on the following papers. And part of the code is based on Megatron-LM and TAPE.
Evaluating Protein Transfer Learning with TAPE
@article{DBLP:journals/corr/abs-1909-08053,
author = {Mohammad Shoeybi and
Mostofa Patwary and
Raul Puri and
Patrick LeGresley and
Jared Casper and
Bryan Catanzaro},
title = {Megatron-LM: Training Multi-Billion Parameter Language Models Using
Model Parallelism},
journal = {CoRR},
volume = {abs/1909.08053},
year = {2019},
url = {http://arxiv.org/abs/1909.08053},
archivePrefix = {arXiv},
eprint = {1909.08053},
timestamp = {Tue, 24 Sep 2019 11:33:51 +0200},
biburl = {https://dblp.org/rec/journals/corr/abs-1909-08053.bib},
bibsource = {dblp computer science bibliography, https://dblp.org}
}
Megatron-LM: Training Multi-Billion Parameter Language Models Using Model Parallelism
@article{DBLP:journals/corr/abs-1906-08230,
author = {Roshan Rao and
Nicholas Bhattacharya and
Neil Thomas and
Yan Duan and
Xi Chen and
John F. Canny and
Pieter Abbeel and
Yun S. Song},
title = {Evaluating Protein Transfer Learning with {TAPE}},
journal = {CoRR},
volume = {abs/1906.08230},
year = {2019},
url = {http://arxiv.org/abs/1906.08230},
archivePrefix = {arXiv},
eprint = {1906.08230},
timestamp = {Sat, 23 Jan 2021 01:20:25 +0100},
biburl = {https://dblp.org/rec/journals/corr/abs-1906-08230.bib},
bibsource = {dblp computer science bibliography, https://dblp.org}
}