c_AMPs-prediction

Prediction of new c_AMPs, from the prediction of sORF to the getting c_AMPs. Scripts.
Trained predictive model.Models.
Test set data under 50AA in length, and also the benchmark dataset.Dataset.

Verified AMPs have been deposited and published in ADP database.
For sequences that have been predicted but not yet verified, please contact Jun Wang(junwang@im.ac.cn) and request a Material Transfer Agreement (MTA) to ensure proper usage.

Installation bert_sklearn

Download and copy bert_sklearn to your python3 site-packages folder.

cd bert-sklearn
pip install .

Please cite: Identification of antimicrobial peptides from the human gut microbiome using deep learning

WLab-Workflow

• 下载资源

git clone https://github.com/BioGavin/c_AMPs-prediction.git
cd c_AMPs-prediction/

• 下载 Bert 模型

从该链接下载 Bert 模型：https://www.dropbox.com/sh/o58xdznyi6ulyc6/AABLckEnxP54j2X7BrGybhyea?dl=0

下载的文件 bert.bin 放至 Models 路径下

校验模型的完整性

cd Models
md5sum -c md5.txt

输出如下结果表示校验通过，否则说明模型文件下载不完整

lstm.h5: OK
att.h5: OK
bert.bin: OK

cd ..  # 返回项目根目录

• 创建环境

conda create -y -n amp_prediction python=3.7 certifi=2022.12.7
conda activate amp_prediction

pip install -r requirement.txt

cd bert_sklearn
pip install .
cd .. && cp -r bert_sklearn/ ~/miniconda3/envs/amp_prediction/lib/python3.7/site-packages/

如上述操作成功完成后，下面仍有包报错问题，参考测试时导出的包列表 requirement_wlabworkflow_reference.txt 进行手动调试

• 序列格式化

# 如果有长度超过300的序列，用如下代码删除
seqkit seq -M 300 -g input.faa -o output.faa
# 如果fasta文件中的序列是分行的，会导致格式化出现问题，用如下代码将序列转换成一行
seqkit seq -w 0 input.faa -o output.faa
# 如果序列中含有X，需要提起手动剔除
seqkit grep -s -p "X" -v input.faa -o output.faa

# 格式化
cd Data
perl ../script/format.pl len5seq.fasta none > len5seq.txt

• Attention 模型预测

python3 ../script/prediction_attention.py len5seq.txt len5seq.att.txt

• LSTM 模型预测

python3 ../script/prediction_lstm.py len5seq.txt len5seq.lstm.txt

• Bert 模型预测

python3 ../script/prediction_bert.py len5seq.fasta len5seq.bert.txt

• 整合结果

python3 ../script/result.py len5seq.att.txt len5seq.lstm.txt len5seq.bert.txt len5seq.fasta len5seq.result.tsv

• 结果可视化

pip install matplotlib
python3 ../customized_scripts/scatterplot3D.py len5seq.att.txt len5seq.lstm.txt len5seq.bert.txt fig.svg