羊村懒王

这个屌丝很懒，什么也没留下！

热门标签

alphafold安装及使用（保姆级教程、conda环境）_alphafold-multimer

作者：羊村懒王 | 2024-06-18 04:09:59

踩

alphafold-multimer

原标题为“alphafold-mutimer的安装及使用”，对一些错误认知进行修改：alphafold-mutimer本身为alphafold的一部分，代码未单独拧出，实际上alphafold-mutimer的在alphafold2.1.2版本开始开发，后期一直随着alphafold更新，所以alphafold-mutimer和alphafold为一体（如图），因此在alphafold官网下载最新版本即可。请使用Linux的客官放心食用，windows客官可作参考。

1.标红为必备软件包

#修改：根据实际情况版本可以变，完全出于有没有报错，，，，

(mm) $ conda list

注：标红色的为必须软件包及对应版本，软件安装要特别注意顺序，会在下一步进行介绍，请徐徐图之。（注意顺序，不然一直重装）

2.系统配置：

安装前需要查看系统配置，然后再搭建conda虚拟环境。

2.1 cuda版本

版本为12.2

nvidia-smi

2.2内存量

free -m

2.3 cudnn版本

版本为89

dpkg -l | grep cudnn

3.conda搭建python3.8的虚拟环境

#修改：python3.9也行

3.1首先需要安装openmm和cudatoolkit，然后检测虚拟环境内的cudatoolkit版本与系统cuda版本是否适配，操作如下：

3.1.1 conda install cudatoolkit==版本号，例如本人安装的为11.8.0 #不能高于系统cuda的版本号，本人使用的系统cuda12.2（用conda search cudatoolkit查找conda仓库中可用的cudatoolkit版本）

3.1.2 conda install -c conda-forge openmm=7.5.1 #alphafold要求openmm版本为7.5.1，不然会报错

3.1.3 开始检测cudatoolkit是否适配，执行：python -m simtk.testInstallation。如果直接执行python -m openmm.testInstallation 可能会报错找不openmm，如果输出以下内容就说明cudatoolkit与系统cuda版本适配：

如果输出内容里出现：CUDA - Error computing forces with CUDA platform，原因在于cudatookit不对。首先使用 nvidia-smi 查看CUDA Version，然后使用 conda install -c conda-forge cudatoolkit= 对应的版本号，就可以解决问题。当然没有cuda也能跑起来，只是费时间。原文链接：（防坑）Alphafold 非docker 安装指南_alphafold安装_Zqinstarking的博客-CSDN博客

3.2 pytorch的安装

#官网Previous PyTorch Versions | PyTorch ，按照cudatoolkit对应版本，安装后检测torch是否支持GPU运行，#相关连接https://blog.csdn.net/weixin_42788078/article/details/103116903，如果支持会返回GPU等相关信息。成功的情况下代码和返回如下（进入python命令行）：


import 	torch
import  time
print(torch.__version__)
print(torch.cuda.is_available())

或者（进入python命令行）


device = torch.device("cuda:0" if (torch.cuda.is_available() and ngpu > 0) else "cpu")
print("驱动为：",device)
print("GPU型号： ",torch.cuda.get_device_name(0))

如果用conda安装后检测不成功，检查（使用conda list）是否为cpu版本的pytorch，在软件包中gpu版本我已标红，如果是cpu版本的，卸了，网上建议用pip重新安装，但是容易断，还是卸了重新用conda安装，大概率会是gpu版本的（╮(╯▽╰)╭）。

#修改：是否为pytorch的cuda版本其实问题不大，alphafold本身是使用TensorFlow连接显卡加速。

3.3 tensorflow安装

安装：

#修改：版本是能使用GPU的就行，该软件十分重要（后期使用显卡加速搭建5个模型，当然也可以使用cpu加速啦，只是比较慢而已，这点上显卡3090平均2小时搭建5个模型，而显卡4090平均1小时搭建5个模型。）

相关安装链接

#安装

conda虚拟环境安装TensorFlow-GPU版本 - 知乎 (zhihu.com)

conda创建Python虚拟环境并安装tensorflow GPU环境_conda search cudnn==7.4-CSDN博客#验证conda创建Python虚拟环境并安装tensorflow GPU环境_conda search cudnn==7.4-CSDN博客


pip install tensorflow-gpu==2.6.0 -i https://pypi.tuna.tsinghua.edu.cn/simple
 
#细节
#请确保numpy版本是1.19.5，否则tensorflow2.6.0与别的1.19.x都很容易打架
#pip install protobuf==3.20.0
#pip install numpy==1.19.5，后面升级为1.23.4最好，不然报错`AttributeError: module 'numpy' has no attribute 'object'`

#检测TensorFlow是否支持GPU，进入python界面，输入如下代码（进入python命令行）


import tensorflow as tf print(tf.test.is_built_with_cuda())
print(tf.test.is_gpu_available()) 
print(tf.config.list_physical_devices('GPU'))

#输出如下即成功

3.4 零零碎碎软件包的安装

如下软件，其他的如果运行有问题再补充。


conda install -y -c conda-forge pdbfixer==1.7 conda install -y -c bioconda hmmer==3.3.2 hhsuite==3.3.0 kalign2==2.04 
pip install absl-py==0.13.0 biopython==1.79 dm-haiku==0.0.4 dm-tree==0.1.6 immutabledict==2.0.0 jax==0.2.14 ml-collections==0.1.0 scipy==1.7.0 pandas==1.3.4
pip install --upgrade jax==0.2.14 jaxlib==0.1.69+cuda111 -f https://storage.googleapis.com/jax-releases/jax_cuda_releases.html

3.5 jax的安装

#JAX: 库安装和GPU使用，解决不能识别gpu问题_jax找不到gpu-CSDN博客


pip install --upgrade jax==0.4.23 jaxlib==0.4.19+cuda11.cudnn86 -f https://storage.googleapis.com/jax-releases/jax_cuda_releases.html
 
#注意根据自己的系统配置安装对应版本，不然又报错，看代码上面的链接，，，
#注意如果jax版本过低则会有很多库无法导出

需要测试

#测试jax是否支持GPU运行，输入如下代码（进入python命令行）：


import torch, jax;
print(torch.cuda.is_available)
print(jax.devices()[0])
gpu:0

#输出如下即成功（进入python命令行）


import torch, jax;
print(torch.cuda.is_available)
<function is_available at 0x7f487840b280>
print(jax.devices()[0])
gpu:0
gpu:0

#测试jax.numpy是否支持GPU运行，输入如下代码（进入python命令行）：


import jax.numpy as np
 
from jax import random

#输出如下即成功

4.alphafold-mutimer使用

注意：在安装过alphafold，数据库已全部下载的前提下进行跟新。

alphafold-mutimer需要用到的数据库在alphafold官网https://github.com/google-deepmind/alphafold下载也可，用最新数据库就行，把alphafold官网上的数据库全部下载下来就可以了，后面只需更改更新模型的参数文件，下一步进行介绍。

4.1更新模型参数文件

在https://github.com/jcheongs/alphafold-multimer或https://github.com/google-deepmind/alphafold下载模型参数，如下：

#下载模型使用参数：wget https://storage.googleapis.com/alphafold/alphafold_params_2022-12-06.tar

#解压后

放入数据库储存路径，如下：params文件内就有模型的参数

4.2 alphafold-mutimer的run_alphafold.py获得

#修改：直接使用alphafold最新版的run_alphafold.py即可

4.3 验证是否能运行alphafold-mutimer

#输入代码：

python run_alphafold.py --help

#输出如下即代表成功：


2023-09-11 15:12:48.147643: I tensorflow/stream_executor/platform/default/dso_loader.cc:53] Successfully opened dynamic library libcudart.so.11.0
2023-09-11 15:12:48.588580: I tensorflow/stream_executor/platform/default/dso_loader.cc:53] Successfully opened dynamic library libcuda.so.1
2023-09-11 15:12:48.607655: I tensorflow/stream_executor/cuda/cuda_gpu_executor.cc:937] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero
2023-09-11 15:12:48.607793: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1733] Found device 0 with properties: 
pciBusID: 0000:01:00.0 name: NVIDIA GeForce RTX 3090 computeCapability: 8.6
coreClock: 1.695GHz coreCount: 82 deviceMemorySize: 23.69GiB deviceMemoryBandwidth: 871.81GiB/s
2023-09-11 15:12:48.607852: I tensorflow/stream_executor/platform/default/dso_loader.cc:53] Successfully opened dynamic library libcudart.so.11.0
2023-09-11 15:12:48.621122: I tensorflow/stream_executor/platform/default/dso_loader.cc:53] Successfully opened dynamic library libcublas.so.11
2023-09-11 15:12:48.621198: I tensorflow/stream_executor/platform/default/dso_loader.cc:53] Successfully opened dynamic library libcublasLt.so.11
2023-09-11 15:12:48.623171: I tensorflow/stream_executor/platform/default/dso_loader.cc:53] Successfully opened dynamic library libcufft.so.10
2023-09-11 15:12:48.623382: I tensorflow/stream_executor/platform/default/dso_loader.cc:53] Successfully opened dynamic library libcurand.so.10
2023-09-11 15:12:48.623839: I tensorflow/stream_executor/platform/default/dso_loader.cc:53] Successfully opened dynamic library libcusolver.so.11
2023-09-11 15:12:48.624399: I tensorflow/stream_executor/platform/default/dso_loader.cc:53] Successfully opened dynamic library libcusparse.so.11
2023-09-11 15:12:48.624489: I tensorflow/stream_executor/platform/default/dso_loader.cc:53] Successfully opened dynamic library libcudnn.so.8
2023-09-11 15:12:48.624560: I tensorflow/stream_executor/cuda/cuda_gpu_executor.cc:937] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero
2023-09-11 15:12:48.624663: I tensorflow/stream_executor/cuda/cuda_gpu_executor.cc:937] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero
2023-09-11 15:12:48.624729: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1871] Adding visible gpu devices: 0
2023-09-11 15:12:48.645313: I tensorflow/core/platform/cpu_feature_guard.cc:142] This TensorFlow binary is optimized with oneAPI Deep Neural Network Library (oneDNN) to use the following CPU instructions in performance-critical operations:  AVX2 FMA
To enable them in other operations, rebuild TensorFlow with the appropriate compiler flags.
2023-09-11 15:12:48.645752: I tensorflow/stream_executor/cuda/cuda_gpu_executor.cc:937] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero
2023-09-11 15:12:48.645845: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1733] Found device 0 with properties: 
pciBusID: 0000:01:00.0 name: NVIDIA GeForce RTX 3090 computeCapability: 8.6
coreClock: 1.695GHz coreCount: 82 deviceMemorySize: 23.69GiB deviceMemoryBandwidth: 871.81GiB/s
2023-09-11 15:12:48.645917: I tensorflow/stream_executor/cuda/cuda_gpu_executor.cc:937] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero
2023-09-11 15:12:48.645984: I tensorflow/stream_executor/cuda/cuda_gpu_executor.cc:937] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero
2023-09-11 15:12:48.646039: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1871] Adding visible gpu devices: 0
2023-09-11 15:12:48.646075: I tensorflow/stream_executor/platform/default/dso_loader.cc:53] Successfully opened dynamic library libcudart.so.11.0
2023-09-11 15:12:48.783877: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1258] Device interconnect StreamExecutor with strength 1 edge matrix:
2023-09-11 15:12:48.783899: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1264]      0 
2023-09-11 15:12:48.783902: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1277] 0:   N 
2023-09-11 15:12:48.783980: I tensorflow/stream_executor/cuda/cuda_gpu_executor.cc:937] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero
2023-09-11 15:12:48.784044: I tensorflow/stream_executor/cuda/cuda_gpu_executor.cc:937] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero
2023-09-11 15:12:48.784090: I tensorflow/stream_executor/cuda/cuda_gpu_executor.cc:937] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero
2023-09-11 15:12:48.784141: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1418] Created TensorFlow device (/job:localhost/replica:0/task:0/device:GPU:0 with 556 MB memory) -> physical GPU (device: 0, name: NVIDIA GeForce RTX 3090, pci bus id: 0000:01:00.0, compute capability: 8.6)
Full AlphaFold protein structure prediction script.
flags:
 
run_alphafold.py:
  --[no]benchmark: Run multiple JAX model evaluations to obtain a timing that excludes the compilation time, which should be more indicative of the time required for inferencing many
    proteins.
    (default: 'false')
  --bfd_database_path: Path to the BFD database for use by HHblits.
  --data_dir: Path to directory of supporting data.
  --db_preset: <full_dbs|reduced_dbs>: Choose preset MSA database configuration - smaller genetic database config (reduced_dbs) or full genetic database config  (full_dbs)
    (default: 'full_dbs')
  --fasta_paths: Paths to FASTA files, each containing a prediction target that will be folded one after another. If a FASTA file contains multiple sequences, then it will be folded as
    a multimer. Paths should be separated by commas. All FASTA paths must have a unique basename as the basename is used to name the output directories for each prediction.
    (a comma separated list)
  --hhblits_binary_path: Path to the HHblits executable.
    (default: '/home/wangyun/miniconda3/envs/mm/bin/hhblits')
  --hhsearch_binary_path: Path to the HHsearch executable.
    (default: '/home/wangyun/miniconda3/envs/mm/bin/hhsearch')
  --hmmbuild_binary_path: Path to the hmmbuild executable.
    (default: '/home/wangyun/miniconda3/envs/mm/bin/hmmbuild')
  --hmmsearch_binary_path: Path to the hmmsearch executable.
    (default: '/home/wangyun/miniconda3/envs/mm/bin/hmmsearch')
  --is_prokaryote_list: Optional for multimer system, not used by the single chain system. This list should contain a boolean for each fasta specifying true where the target complex is
    from a prokaryote, and false where it is not, or where the origin is unknown. These values determine the pairing method for the MSA.
    (a comma separated list)
  --jackhmmer_binary_path: Path to the JackHMMER executable.
    (default: '/home/wangyun/miniconda3/envs/mm/bin/jackhmmer')
  --kalign_binary_path: Path to the Kalign executable.
    (default: '/home/wangyun/miniconda3/envs/mm/bin/kalign')
  --max_template_date: Maximum template release date to consider. Important if folding historical test sets.
  --mgnify_database_path: Path to the MGnify database for use by JackHMMER.
  --model_preset: <monomer|monomer_casp14|monomer_ptm|multimer>: Choose preset model configuration - the monomer model, the monomer model with extra ensembling, monomer model with pTM
    head, or multimer model
    (default: 'monomer')
  --obsolete_pdbs_path: Path to file containing a mapping from obsolete PDB IDs to the PDB IDs of their replacements.
  --output_dir: Path to a directory that will store the results.
  --pdb70_database_path: Path to the PDB70 database for use by HHsearch.
  --pdb_seqres_database_path: Path to the PDB seqres database for use by hmmsearch.
  --random_seed: The random seed for the data pipeline. By default, this is randomly generated. Note that even if this is set, Alphafold may still not be deterministic, because
    processes like GPU inference are nondeterministic.
    (an integer)
  --[no]run_relax: Whether to run the final relaxation step on the predicted models. Turning relax off might result in predictions with distracting stereochemical violations but might
    help in case you are having issues with the relaxation stage.
    (default: 'true')
  --small_bfd_database_path: Path to the small version of BFD used with the "reduced_dbs" preset.
  --template_mmcif_dir: Path to a directory with template mmCIF structures, each named <pdb_id>.cif
  --uniclust30_database_path: Path to the Uniclust30 database for use by HHblits.
  --uniprot_database_path: Path to the Uniprot database for use by JackHMMer.
  --uniref90_database_path: Path to the Uniref90 database for use by JackHMMER.
  --[no]use_gpu_relax: Whether to relax on GPU. Relax on GPU can be much faster than CPU, so it is recommended to enable if possible. GPUs must be available if this setting is enabled.
  --[no]use_precomputed_msas: Whether to read MSAs that have been written to disk instead of running the MSA tools. The MSA files are looked up in the output directory, so it must stay
    the same between multiple runs that are to reuse the MSAs. WARNING: This will not check if the sequence, database or configuration have changed.
    (default: 'false')
 
Try --helpfull to get a list of all flags.

5. 如何使用alphafold-mutimer参数

#修改：注：alphafold跟新后没有该参数（--is_prokaryote_list=false），该参数在alphafold2.1.2版本存在，这里就不要管了。

5.1使用大的数据库


python run_alphafold.py 
--data_dir=/data/luping/alphaFold/alphaFold2/alphafold/scripts/AlphaFoldDataRes#数据库所在的大目录 
--uniref90_database_path=/data/luping/alphaFold/alphaFold2/alphafold/scripts/AlphaFoldDataRes/uniref90/uniref90.fasta#（数据库大目录下uniref90的目录） 
--mgnify_database_path=/data/luping/alphaFold/alphaFold2/alphafold/scripts/AlphaFoldDataRes/mgnify/mgy_clusters_2022_05.fa#（数据库大目录下mgnify的目录） 
--template_mmcif_dir=/data/luping/alphaFold/alphaFold2/alphafold/scripts/AlphaFoldDataRes/pdb_mmcif/mmcif_files#（数据库大目录下mmcif的目录） 
--max_template_date=2021-11-01 
--obsolete_pdbs_path=/data/luping/alphaFold/alphaFold2/alphafold/scripts/AlphaFoldDataRes/pdb_mmcif/obsolete.dat#（数据库大目录下obsolete的目录） 
--bfd_database_path=/data/luping/alphaFold/alphaFold2/alphafold/scripts/AlphaFoldDataRes/bfd/bfd_metaclust_clu_complete_id30_c90_final_seq.sorted_opt#（数据库大目录下bfd的目录） 
--model_preset=multimer#（试验项目为多聚体） 
--uniprot_database_path=/data/luping/alphaFold/alphaFold2/alphafold/scripts/AlphaFoldDataRes/uniprot/uniprot.fasta#（数据库大目录下uniprot的目录） 
--pdb_seqres_database_path=/data/luping/alphaFold/alphaFold2/alphafold/scripts/AlphaFoldDataRes/pdb_seqres/pdb_seqres.txt#（数据库大目录下pdb_seqres的目录） 
--db_preset=full_dbs#（试验以大数据库为仓库筛选模板） 
--uniclust30_database_path=/data/luping/alphaFold/alphaFold2/alphafold/scripts/AlphaFoldDataRes/uniref30/UniRef30_2021_03#（数据库大目录下UniRef30的目录）   
--fasta_paths=/home/wangyun/pre/proin/mutimer1 #（输入的fasta文件，为了准确性内含2条及两条以上的序列）
--is_prokaryote_list=false#（试验序列为真核生物选true，否则为false）
--output_dir=/home/wangyun/pre/outbig-m #（输出结果文件的目录）
--use_gpu_relax=true#（是否使用GPU进行松弛处理，处理后模型打分会更高） 
--benchmark=true#（加速？）
--use_precomputed_msas=true#（是否使用原先生成的msas文件，测试同一个fasta文件序列各种参数的时候可用）

5.2使用小的数据库


python run_alphafold.py 
--data_dir=/data/luping/alphaFold/alphaFold2/alphafold/scripts/AlphaFoldDataRes#（参考使用大数据库） 
--uniref90_database_path=/data/luping/alphaFold/alphaFold2/alphafold/scripts/AlphaFoldDataRes/uniref90/uniref90.fasta#（参考使用大数据库） 
--mgnify_database_path=/data/luping/alphaFold/alphaFold2/alphafold/scripts/AlphaFoldDataRes/mgnify/mgy_clusters_2022_05.fa#（参考使用大数据库） 
--template_mmcif_dir=/data/luping/alphaFold/alphaFold2/alphafold/scripts/AlphaFoldDataRes/pdb_mmcif/mmcif_files#（参考使用大数据库） 
--is_prokaryote_list=false#（参考使用大数据库） 
--max_template_date=2021-11-01#（参考使用大数据库） 
--obsolete_pdbs_path=/data/luping/alphaFold/alphaFold2/alphafold/scripts/AlphaFoldDataRes/pdb_mmcif/obsolete.dat#（参考使用大数据库） 
--model_preset=multimer#（参考使用大数据库） 
--uniprot_database_path=/data/luping/alphaFold/alphaFold2/alphafold/scripts/AlphaFoldDataRes/uniprot/uniprot.fasta#（参考使用大数据库） 
--pdb_seqres_database_path=/data/luping/alphaFold/alphaFold2/alphafold/scripts/AlphaFoldDataRes/pdb_seqres/pdb_seqres.txt#（参考使用大数据库） 
--db_preset=reduced_dbs#（使用小数据库） 
--small_bfd_database_path=/data/luping/alphaFold/alphaFold2/alphafold/scripts/AlphaFoldData/small_bfd/bfd-first_non_consensus_sequences.fasta#（参考使用大数据库） 
--fasta_paths=/home/wangyun/pre/proin/mutimer1#（参考使用大数据库） 
--output_dir=/home/wangyun/pre/out-m#（参考使用大数据库） 
--use_gpu_relax=true#（参考使用大数据库） 
--benchmark=true --use_precomputed_msas=true#（参考使用大数据库）

6.结果

正常情况下会有5个模型输出，ranked_0.pdb打分最高，在ranking_debug.json中可得到 "iptm+ptm"分数（model confidence（DockQ） = 0.8 · ipTM + 0.2 · pTM 成功（DockQ ≥ 0.23），中等准确度(DockQ [3]≥0.49) ，高准确度(DockQ ≥0.8) ，该分数高低代表模型的可信度。参考文献Protein complex prediction with AlphaFold-Multimer (deepmind.com)）。之后可以下载ChimeraX软件UCSF ChimeraX Home Page后打开.pdb图像文件。

#图像

7.意外

alphafold属于机器训练，出错再所难免，我遇到这个问题：5个模型打分天差地别，建议重复预测其模型。出错率7%吧，如果是各位需要深入研究的蛋白，建议取打分靠前的蛋白，两次预测其结构模型（在5个模型打分相差巨大的情况下）。最后祝各位科研顺利！

本人属于跨界接触深度学习，不当之处还望各位大佬指出，您们的每一次严厉指出是我进步的推力，谢谢各位。

声明：本文内容由网友自发贡献，转载请注明出处：【wpsshop博客】