弃坑（GraphEC的安装）ESMfold安装

GraphEC
Accurately predicting enzyme functions through geometric graph learning on ESMFold-predicted structures

官方教程一如既往的不好用

创建一个新环境，安装依赖

conda create -n GraphEC python=3.8.16
conda activate GraphEC
#注意这里NVCC不能是12.5


conda install -y -c conda-forge numpy=1.24.3 
conda install -y -c conda-forge biopython=1.81
conda install -y -c conda-forge debugpy=1.6.7
conda install -y -c conda-forge decorator=5.1.1
conda install -y -c conda-forge filelock=3.12.1
conda install -y -c conda-forge gmp=6.2.1
conda install -y -c conda-forge idna=3.4
conda install -y -c conda-forge scipy=1.10.1
conda install -y -c conda-forge six=1.16.0


pip install torch-geometric
#pip install biopython==1.81

#安装ProtTrans
pip install torch
#pip install transformers
conda install conda-forge::transformers
pip install sentencepiece
pip install protobuf
pip install scikit-learn
pip install fair-esm
conda install conda-forge::torch-scattery



git clone https://github.com/biomed-AI/GraphEC.git
chmod -R 755 ./EC_number/tools/
chmod +x ./

可能有用的wsl 12.4安装参考：https://blog.csdn.net/qq_74025626/article/details/139048062

检查一下

import numpy as np
import torch
import torch_geometric
import biopython
import debugpy
import decorator
import filelock
import gmpy2
import idna
import IPython
import scipy
import six

print("numpy version:", np.__version__)
print("pytorch version:", torch.__version__)
print("torch-geometric version:", torch_geometric.__version__)
print("biopython version:", biopython.__version__)
print("debugpy version:", debugpy.__version__)
print("decorator version:", decorator.__version__)
print("filelock version:", filelock.__version__)
print("gmpy2 version:", gmpy2.__version__)
print("idna version:", idna.__version__)
print("IPython version:", IPython.__version__)
print("scipy version:", scipy.__version__)
print("six version:", six.__version__)

#检查一下torch
torch.cuda.is_available()

下载模型权重

sudo mkdir -p /home/songyd/software
cd /home/songyd/software
sudo 你的地址/mv prot_t5_xl_uniref50.zip /home/songyd/software

sudo unzip prot_t5_xl_uniref50.zip
sudo mv /home/songyd/software/prot_t5_xl_uniref50 /home/songyd/software/Prot-T5-XL-U50

跑一下测试

bash run.sh EC_number ./Data/fasta/EC_number.fasta --gpu 0

到这里应该没问题了，但是自己用要安装ESMfold，不知道为啥我的esm报错了（40系列显卡WSL默认装的是cuda 12.5，很烦）

参考下别人的教程重新安装环境

参考：https://zhuanlan.zhihu.com/p/680833258
https://github.com/RazzyChen/One-command-install-ESMfold/tree/main

pip install transformer
pip install torch-geometric
pip install torch-scatter
pip install torch-geometric torch-cluster
pip install sentencepiece
pip install protobuf