DeepRank

安装

你可能需要手动安装pytorch geometric

• pytorch_geometric : https://github.com/rusty1s/pytorch_geometric

当所有依赖项安装完毕后，克隆仓库并使用以下命令安装：

pip install -e ./

文档可在此处找到： https://deeprank-gnn.readthedocs.io/

生成图

使用GenGraph.py脚本，将存储在data/pdb/和data/pssm/中的所有pdb/pssm的图/线图生成。这将生成包含不同构象图的hdf5文件graph_residue.hdf5。

from GraphGen import GraphHDF5

pdb_path = './data/pdb'
pssm_path = './data/pssm'
ref = './data/ref'

GraphHDF5(pdb_path=pdb_path,ref_path=ref,pssm_path=pssm_path,
	      graph_type='residue',outfile='graph_residue.hdf5')

图交互网络

使用图交互网络相当简单

from deeprank_gnn.NeuralNet import NeuralNet
from deeprank_gnn.ginet import GINet

database = './hdf5/1ACB_residue.hdf5'

NN = NeuralNet(database, GINet,
               node_feature=['type', 'polarity', 'bsa',
                             'depth', 'hse', 'ic', 'pssm'],
               edge_feature=['dist'],
               target='irmsd',
               index=range(400),
               batch_size=64,
               percent=[0.8, 0.2])

NN.train(nepoch=250, validate=False)
NN.plot_scatter()

自定义CNN

也可以定义新的网络架构并指定在训练期间使用的损失和优化器。

def normalized_cut_2d(edge_index, pos):
    row, col = edge_index
    edge_attr = torch.norm(pos[row] - pos[col], p=2, dim=1)
    return normalized_cut(edge_index, edge_attr, num_nodes=pos.size(0))


class CustomNet(torch.nn.Module):
    def __init__(self):
        super(Net, self).__init__()
        self.conv1 = SplineConv(d.num_features, 32, dim=2, kernel_size=5)
        self.conv2 = SplineConv(32, 64, dim=2, kernel_size=5)
        self.fc1 = torch.nn.Linear(64, 128)
        self.fc2 = torch.nn.Linear(128, 1)

    def forward(self, data):
        data.x = F.elu(self.conv1(data.x, data.edge_index, data.edge_attr))
        weight = normalized_cut_2d(data.edge_index, data.pos)
        cluster = graclus(data.edge_index, weight)
        data = max_pool(cluster, data)

        data.x = F.elu(self.conv2(data.x, data.edge_index, data.edge_attr))
        weight = normalized_cut_2d(data.edge_index, data.pos)
        cluster = graclus(data.edge_index, weight)
        x, batch = max_pool_x(cluster, data.x, data.batch)

        x = scatter_mean(x, batch, dim=0)
        x = F.elu(self.fc1(x))
        x = F.dropout(x, training=self.training)
        return F.log_softmax(self.fc2(x), dim=1)


device = torch.device('cuda' if torch.cuda.is_available() else 'cpu')
model = NeuralNet(database, CustomNet,
               node_feature=['type', 'polarity', 'bsa',
                             'depth', 'hse', 'ic', 'pssm'],
               edge_feature=['dist'],
               target='irmsd',
               index=range(400),
               batch_size=64,
               percent=[0.8, 0.2])
model.optimizer = torch.optim.Adam(model.parameters(), lr=0.01)
model.loss = MSELoss()

model.train(nepoch=50)

h5x支持

安装h5xplorer（https://github.com/DeepRank/h5xplorer）后，可以执行python文件deeprank_gnn/h5x/h5x.py来探索DeepRank-GNN使用的连接图。上下文菜单（在结构名称上右键单击）允许使用下面的plotly自动绘制图表。