AI 药物发现的完整计算流程代码教程

itwangyang520

730人浏览 · 2026-03-11 16:49:27

itwangyang520 · 2026-03-11 16:49:27 发布

AI 药物发现的完整计算流程代码教程

这些技术构成了当前 AI 药物发现的完整计算流程。例如 AlphaFold 系列模型已经成为蛋白结构预测与药物发现的重要基础工具。 (https://github.com/google-deepmind/alphafold?utm_source=chatgpt.com)

AlphaFold 生态
Protein LLM
Diffusion drug design
GNN drug discovery
Docking / virtual screening
Drug repurposing
De-novo drug design
Pharma AI pipelines

AIDD GitHub 超级地图

AlphaFold / Protein Structure Prediction（15）

项目	GitHub
AlphaFold	https://github.com/google-deepmind/alphafold
AlphaFold3	https://github.com/google-deepmind/alphafold3
OpenFold	https://github.com/aqlaboratory/openfold
FastFold	https://github.com/hpcaitech/FastFold
ColabFold	https://github.com/sokrypton/ColabFold
RoseTTAFold	https://github.com/RosettaCommons/RoseTTAFold
RoseTTAFold-All-Atom	https://github.com/RosettaCommons/RoseTTAFold-All-Atom
ProteinMPNN	https://github.com/dauparas/ProteinMPNN
AlphaPulldown	https://github.com/KosinskiLab/AlphaPulldown
HelixFold	https://github.com/PaddlePaddle/PaddleHelix
MiniAlphaFold	https://github.com/sachiin044/AlphaFold
AlphaFind	https://github.com/Coda-Research-Group/AlphaFind
AlphaSMILES	https://github.com/grogdrinker/AlphaSMILES
Boltz	https://github.com/recursionpharma/boltz
Boltz-2	https://github.com/recursionpharma/boltz

Protein Language Models（Protein LLM）（12）

项目	GitHub
ESM	https://github.com/facebookresearch/esm
ProtTrans	https://github.com/agemagician/ProtTrans
ProteinBERT	https://github.com/nadavbra/protein_bert
ProGen	https://github.com/salesforce/progen
ProGen2	https://github.com/salesforce/progen2
ProteinLM	https://github.com/IBM/protein-lm
ProtGPT2	https://github.com/agemagician/ProtGPT2
RITA	https://github.com/lightonai/RITA
OmegaFold	https://github.com/HeliXonProtein/OmegaFold
Ankh	https://github.com/agemagician/Ankh
ProteinGym	https://github.com/OATML-Markslab/ProteinGym
SaProt	https://github.com/westlake-repl/SaProt

Diffusion Drug Design（扩散模型）（15）

项目	GitHub
DiffDock	https://github.com/gcorso/DiffDock
DiffDock-PP	https://github.com/ketatam/DiffDock-PP
RFdiffusion	https://github.com/RosettaCommons/RFdiffusion
DecompDiff	https://github.com/bytedance/DecompDiff
TargetDiff	https://github.com/guanjq/targetdiff
GeoDiff	https://github.com/MinkaiXu/GeoDiff
Pocket2Mol	https://github.com/pengxingang/Pocket2Mol
MolDiff	https://github.com/guanjq/mol-diffusion
BindDM	https://github.com/YangLing0818/BindDM
DiffSBDD	https://github.com/arneschneuing/DiffSBDD
GraphDF	https://github.com/divelab/DIG
DrugDiff	https://github.com/yangkevin2/drugdiff
GCDM	https://github.com/biomed-AI/GCDM
EquivariantDiffusion	https://github.com/ehoogeboom/e3_diffusion
DeNovoDiffusion	https://github.com/biomed-AI/DeNovoDiffusion

扩散模型目前是 AI药物设计最活跃研究方向之一，用于生成新的候选分子结构。 (arXiv)

GNN Drug Discovery（15）

项目	GitHub
DeepChem	https://github.com/deepchem/deepchem
DeepPurpose	https://github.com/kexinhuang12345/DeepPurpose
Chemprop	https://github.com/chemprop/chemprop
DeepMol	https://github.com/BioSystemsUM/DeepMol
TorchDrug	https://github.com/DeepGraphLearning/torchdrug
DGL-LifeSci	https://github.com/awslabs/dgl-lifesci
MolCLR	https://github.com/yuyangw/MolCLR
Uni-Mol	https://github.com/dptech-corp/Uni-Mol
GraphDTA	https://github.com/thinng/GraphDTA
DeepDTA	https://github.com/hkmztrk/DeepDTA
MolTrans	https://github.com/kexinhuang12345/MolTrans
CPI-GNN	https://github.com/masashitsubaki/CPI-GNN
NeoDTI	https://github.com/luoyunan/NeoDTI
DTINet	https://github.com/luoyunan/DTINet
MolGraph	https://github.com/chemprop/chemprop

Docking / Virtual Screening（12）

项目	GitHub
AutoDock4	https://github.com/ccsb-scripps/AutoDock4
AutoDock Vina	https://github.com/ccsb-scripps/AutoDock-Vina
GNINA	https://github.com/gnina/gnina
FlexAID	https://github.com/NRGlab/FlexAID
rDock	https://github.com/CBDD/rDock
QuickVina2	https://github.com/QVina/qvina
Vina-GPU	https://github.com/DeltaGroupNJUPT/Vina-GPU
Smina	https://github.com/mwojcikowski/smina
OpenVS	https://github.com/VirtualFlow/OpenVS
VirtualFlow	https://github.com/VirtualFlow/VFVS
DeepDock	https://github.com/OptiMaL-PSE-Lab/DeepDock
DockStream	https://github.com/MolecularAI/DockStream

Docking 用于预测分子与蛋白结合方式，是虚拟筛选的重要环节。 (Wikipedia)

De-novo Drug Design（生成模型）（15）

项目	GitHub
REINVENT	https://github.com/MolecularAI/Reinvent
MolGPT	https://github.com/devalab/molgpt
DrugEx	https://github.com/CDDLeiden/DrugEx
MolGAN	https://github.com/nicola-decao/MolGAN
GraphAF	https://github.com/DeepGraphLearning/GraphAF
JT-VAE	https://github.com/wengong-jin/icml18-jtnn
GraphEBM	https://github.com/DeepGraphLearning/GraphEBM
GFlowNet	https://github.com/GFNOrg/gflownet
PCMol	https://github.com/CDDLeiden/PCMol
LIMO	https://github.com/Rose-STL-Lab/LIMO
ACEGEN	https://github.com/acellera/acegen-open
MolRL-MGPT	https://github.com/HXYfighter/MolRL-MGPT
SMILES-GPT	https://github.com/szczurek-lab/smiles-gpt
ChemGPT	https://github.com/devalab/ChemGPT
GraphInvent	https://github.com/MolecularAI/GraphInvent

Drug Repurposing / DTI（8）

项目	GitHub
DeepPurpose	https://github.com/kexinhuang12345/DeepPurpose
MolTrans	https://github.com/kexinhuang12345/MolTrans
TransformerCPI	https://github.com/lifanchen-simm/transformerCPI
DrugBAN	https://github.com/peizhenbai/DrugBAN
DeepAffinity	https://github.com/Shen-Lab/DeepAffinity
CPI-GNN	https://github.com/masashitsubaki/CPI-GNN
DTINet	https://github.com/luoyunan/DTINet
NeoDTI	https://github.com/luoyunan/NeoDTI

Pharma AI Pipelines（10）

项目	GitHub
AISCULAPIUS	https://github.com/atwalling/AISCULAPIUS
DeepPurpose-MD-Discovery	https://github.com/BioMolDynamics/DeepPurpose-MD-Discovery
VirtualFlow	https://github.com/VirtualFlow/VFVS
DrugChat	https://github.com/UCSD-AI4H/drugchat
RDKit	https://github.com/rdkit/rdkit
OpenBabel	https://github.com/openbabel/openbabel
Avogadro	https://github.com/OpenChemistry/avogadro
DeepChem Pipelines	https://github.com/deepchem/deepchem
DrugAI awesome list	https://github.com/yataobian/awesome-DrugAI
AI Protein Design list	https://github.com/opendilab/awesome-AI-based-protein-design

RDKit 和 OpenBabel 是药物化学数据处理的重要基础工具。 (Wikipedia)

AIDD GitHub 项目统计

类别	项目数量
AlphaFold生态	15
Protein LLM	12
Diffusion药物生成	15
GNN药物发现	15
Docking	12
De-novo设计	15
Drug Repurposing	8
Pharma pipeline	10

总计：102 个项目

AIDD 技术地图（完整 pipeline）

Protein sequence
        │
        ▼
Protein LLM
(ESM / ProtTrans)
        │
        ▼
Structure prediction
(AlphaFold / RoseTTAFold)
        │
        ▼
Binding site detection
        │
        ▼
Docking / Screening
(AutoDock / DiffDock)
        │
        ▼
Molecule generation
(MolGAN / Diffusion)
        │
        ▼
Property prediction
(DeepChem / Chemprop)
        │
        ▼
Lead optimization
(REINVENT / DrugEx)

最推荐学习的 12 个核心项目

DeepChem
DeepPurpose
Chemprop
TorchDrug
UniMol
DiffDock
RFdiffusion
MolGAN
AutoDock Vina
REINVENT
ProteinMPNN
AlphaFold

MCP技术社区

欢迎加入 MCP 技术社区！与志同道合者携手前行，一同解锁 MCP 技术的无限可能！

更多推荐

Google能索引，AI却抓不到

很多技术团队有个错觉：只要Google能正常收录，内容就没问题。但现实是，你的页面在Google搜索结果里好好的，在ChatGPT、Perplexity、Claude的答案里却像从未存在过一样。

MCP技术社区

Windows 下部署 New API 中转站全过程实录（含 6 个真实踩坑）

MCP技术社区

初识智能体

在过去两年中，几乎所有人都体验过大语言模型（如 ChatGPT、Claude、DeepSeek）的强大能力：写周报、润色邮件、解答编程疑难。被动响应：你问一句，它答一句。如果你的 Prompt 没有写清楚约束条件，它无法主动追问或修正目标。知行脱节：大模型能用流畅的文字告诉你“如何部署一个高可用 K8s 集群”，但它无法替你登录服务器敲下命令、查看报错日志并自动回滚。信息孤岛：大模型的知识止步于预