序号 | 名称 | 来源/下载地址 | 相关文献 | 安装步骤 | 测试 |
---|---|---|---|---|---|
01 | ShapeProtrudeDist | rdkit | - | - | ✓ |
02 | ShapeTanimotoDist | rdkit | - | - | ✓ |
03 | ShapeTverskyIndex | rdkit | - | - | ✓ |
04 | SC score | rdkit | @Yang2020 | - | ✓ |
05 | Gobbi_Pharm2D | rdkit | - | - | ✓ |
06 | USR | oddt | @Ballester2007 | - | ✓ |
07 | USRCAT | oddt | @Schreyer2012 | - | ✓ |
08 | Electroshape | oddt | @Armstrong2010 | - | ✓ |
09 | ACPC | riken | @Berenger2014 | Installation | ✓ |
10 | ESP-Sim | pip | @Bolcalto2021 | - | ✓ |
11 | PAPER | SimTK | @Haque2010 | software | ✗ |
12 | MolShaCS | code.google | @VazdeLima2013 | - | ✗ |
13 | 🌟SHAFTS | lilab | @Liu2011 | software | ✗ |
14 | ShaEP | mivainio | @Vainio2009 | - | ✗ |
15 | SimG | lilab | @Cai2013 | software | ✗ |
16 | EGNN | git | @Satorras2022 | usage | ✗ |
- | ✗ |
- 相关综述:
- [@Bero2017] Similarity Measure for Molecular Structure: A Brief Review
- [@Kumar2018] Advances in the Development of Shape Similarity Methods and Their Application in Drug Discovery
- [@Jiang2021a] A comprehensive comparative assessment of 3D molecular similarity tools in ligand-based virtual screening
- Install Opam
sudo apt install software-properties-common sudo add-apt-repository ppa:avsm/ppa # may get message that `ERROR: '~avsm' user or team does not exist`, please try again sudo apt update sudo apt install gnuplot-x11 autoconf opam make gcc patch
- Initialization Opam
rm -rf ~/.opam opam init --disable-sandboxin # for LXC container
- Install ACPC
opam install ACPC # try until prompt change from `Processing 1/1: [default: http]` to `Processing 1/1`:
已下载在molecule-3d-similarity/softwares/目录下
已下载在molecule-3d-similarity/softwares/目录下, 如果出现./Cynthia: No such file or directory
错误,可以尝试安装sudo apt-get install lib32stdc++6
(参考)。
已下载在molecule-3d-similarity/softwares/目录下
.
├── ae_datasets
│ ├── dataloader.py
│ ├── d_creator.py
│ ├── d_selector.py
│ └── __init__.py
├── dude # need to notice
│ ├── data # dude preprocess data
│ │ ├── aa2ar.csv
│ │ ├── abl1.csv
│ │ ├── ace.csv
│ │ ├── aces.csv
...
│ │ └── xiap.csv
│ ├── data.py # load data utils
│ ├── __init__.py
│ ├── models.py # construct models
│ └── utils.py
├── eval.py
├── graph.py
├── LICENSE
├── losess.py
├── main_ae.py
├── main_dude.py # dude main
├── main_nbody.py
├── main_qm9.py
├── models
│ ├── ae.py
│ ├── egnn_clean
│ │ ├── egnn_clean.py
│ │ └── __init__.py
│ ├── egnn.png
│ ├── gcl.py
│ └── __init__.py
├── n_body_system
├── qm9
├── README.md
└── utils.py
bs -> batch size
nn -> node number
nt -> node type
cp -> charge power
pm -> postion number (len(x, y, z) = 3)
-
模型构建参数
名称 含义 类型 取值 备注 1 in_node_nf 输入节点类型数 int 11*(2+1) nf=number of feature;
in_node_nf=nt*(cp+1);2 in_edge_nf 输入边类型数 int 4 (single, double, triple, aromatic) 3 hidden_nf embedding维度 int 128 4 device 模型载入设备 torch.device - 5 n_layers E_GCL数量 int 7 6 attention 注意力机制 bool - -
模型输入参数
名称 含义 形状 关联中间变量 备注 1 h Nodes输入向量 (2, bs*nn, nt*(cp+1)) one_hot, chargs, charge_power,
charge_scale, device输入均为包含两个元素的列表,
第一元素代表ref,第二个元素代表prb2 x Nodes的三维坐标 (2, bs*nn, pm) - 3 edges Edges输入向量 (2, 2, bs*(n^2-3n+1)) - 列表中每一个元素
形如[tensor1, tensor2]这样的方式构建;
代表[rows, cols];
只记录邻接矩阵中的上三角矩阵中的边4 edge_attr 边属性向量 (2, bs*(n^2-3n+1), 5) - 5 node_mask 掩模padding节点 (2, bs, nn) charges 动态padding;
未使用;6 edge_mask 掩模不存在边 (2, bs*(n^2-3n+1), 1) 7 n_nodes 节点数量 (2, 1) 8 label 标签 (1) actives还是decoys
- 数据预处理
使用molecule-3d-similarity/test.ipynb文件中的代码,对DUD-E的数据进行预处理,对每一个靶点得到具有
(name, smiles, label, charges, position, edges)
列的CSV文件。 - DataUp
waiting...
- Dataset
waiting...
- DataLoader
waiting...
-
sudo: add-apt-repository: command not found https://linuxconfig.org/sudo-apt-add-repository-command-not-found
sudo apt install software-properties-common