Ligand screening model construction method and device, a screening method, a device, and a medium

1 . A ligand screening model construction method, which comprises:

obtaining a drug ligand training set, wherein the drug ligand training set includes a drug ligand chemical formula and one or more classification labels;

based on the ligand chemical formula, drawing a ligand graph network, in which atoms are nodes and chemical bonds are edges connecting the nodes;

using a random initialization vector to identify a weight vector of each node in the ligand graph network;

reconstructing each node of the ligand graph network according to a connection relationship of the ligand graph network to obtain a reconstruction graph network, and

repeating the reconstruction step to obtain a reconstruction graph network comprising at least two layers;

performing deep learning on the ligand graph network and the reconstruction graph network comprising at least two layers, according to one or more classification labels; and

constructing a ligand screening model.

2 . The method according to claim 1 , wherein using a random initialization vector to identify the weight vector of each node in the ligand graph network includes:

using at least one of a uniform distribution model and a normal distribution model to select a value of a (0,1) interval to generate the initialization vector; and

assigning the initialization vector to the node as the weight vector.

3 . The method according to claim 1 , wherein reconstructing each node of the ligand graph network according to the connection relationship of the ligand graph network to obtain a reconstruction network includes:

obtaining the weight vector of each node and an associated weight vector of an associated node connected with the node according to the connection relationship of the ligand graph network;

integrating the weight vector and the associated weight vector together to obtain a reconstruction weight vector of the node; and

generating the reconstruction network according to the reconstruction weight vector.

4 . The method according to claim 1 , wherein the ligand graph network and the reconstruction graph network comprising at least two layers are deeply learned according to the classification labels; and further comprising:

constructing a ligand screening model, which includes:

respectively extracting features of the ligand graph network and the reconstruction graph network comprising at least two layers, to obtain hierarchical network features;

according to the classification label, adopting a deep learning network to learn hierarchical network characteristics; and

constructing the ligand screening model.

5 . The method according to claim 4 , wherein extracting features is performed on the ligand graph network and the reconstruction graph network comprising at least two layers respectively to obtain hierarchical network features, including:

performing feature extraction on the ligand graph network and the reconstruction graph network comprising at least two layers respectively through a function to obtain function vector features; and

normalizing the function vector features in probability space to obtain hierarchical network features.

6 . A screening drug ligands method, comprising:

drawing a target ligand graph network based on a ligand chemical formula of the target ligand, in which atoms are nodes and chemical bonds are edges connecting the nodes;

inputting the target ligand graph network into a ligand screening model; and

outputting an analysis result of the target ligand, wherein the ligand screening model is trained by a method comprising:

a ligand screening model construction method, which comprises:

obtaining a drug ligand training set, wherein the drug ligand training set includes a drug ligand chemical formula and one or more classification labels;

based on the ligand chemical formula, drawing the ligand graph network, in which atoms are nodes and chemical bonds are edges connecting the nodes;

using a random initialization vector to identify a weight vector of each node in the ligand graph network;

reconstructing each node of the ligand graph network according to a connection relationship of the ligand graph network to obtain a reconstruction graph network, and repeating the reconstruction step to obtain a reconstruction graph network comprising at least two layers;

performing deep learning on the ligand graph network and the reconstruction graph network comprising at least two layers, according to one or more classification labels; and

constructing the ligand screening model.

7 . A ligand screening model construction device, characterized in that the device comprises:

a training set acquisition module, configured to acquire a drug ligand training set, wherein the drug ligand training set includes a drug ligand chemical formula and a classification label;

a network drawing module is used to draw a ligand graph network based on the ligand chemical formula, in which atoms are nodes and chemical bonds are edges connecting nodes;

a vector identification module configured to identify a weight vector of each node in the ligand graph network by using a random initialization vector;

a reconstruction module used to reconstruct each node of the ligand graph network according to a connection relationship of the ligand graph network to obtain a reconstruction graph network, and repeating the reconstruction steps to obtain a reconstruction graph network comprising at least two layers; and

a model training module used for deeply learning the ligand graph network and the reconstruction graph network comprising at least two layers, according to the classification labels and construct a ligand screening model.

8 . A computer device comprising a memory and a processor, wherein the memory stores a computer program, and is characterized in that when executing the computer program, the processor implements steps of a method comprising:

a ligand screening model construction method, which comprises:

obtaining a drug ligand training set, wherein the drug ligand training set includes a drug ligand chemical formula and one or more classification labels;

based on the ligand chemical formula, drawing a ligand graph network, in which atoms are nodes and chemical bonds are edges connecting the nodes;

using a random initialization vector to identify a weight vector of each node in the ligand graph network;

reconstructing each node of the ligand graph network according to a connection relationship of the ligand graph network to obtain a reconstruction graph network, and

repeating the reconstruction step to obtain a reconstruction graph network comprising at least two layers;

performing deep learning on the ligand graph network and the reconstruction graph network comprising at least two layers, according to one or more classification labels; and

constructing the ligand screening model.

9 . The method according to claim 8 , wherein using a random initialization vector to identify the weight vector of each node in the ligand graph network includes:

using at least one of a uniform distribution model and a normal distribution model to select a value of a (0,1) interval to generate the initialization vector; and

assigning the initialization vector to the node as the weight vector.

10 . The method according to claim 8 , wherein reconstructing each node of the ligand graph network according to the connection relationship of the ligand graph network to obtain a reconstruction network includes:

obtaining the weight vector of each node and an associated weight vector of an associated node connected with the node according to the connection relationship of the ligand graph network;

integrating the weight vector and the associated weight vector together to obtain a reconstruction weight vector of the node; and

generating the reconstruction network according to the reconstruction weight vector.

11 . The method according to claim 8 , wherein the ligand graph network and the reconstruction graph network comprising at least two layers are deeply learned according to the classification labels; and further comprising:

constructing a ligand screening model, which includes:

respectively extracting features of the ligand graph network and the reconstruction graph network comprising at least two layers, to obtain hierarchical network features;

according to the classification label, adopting a deep learning network to learn hierarchical network characteristics; and

constructing the ligand screening model.

12 . The method according to claim 11 , wherein extracting features is performed on the ligand graph network and the reconstruction graph network comprising at least two layers respectively to obtain hierarchical network features, including:

performing feature extraction on the ligand graph network and the reconstruction graph network comprising at least two layers respectively through a function to obtain function vector features; and

normalizing the function vector features in probability space to obtain hierarchical network features.

13 . A non-transitory computer readable storage medium on which a computer program is stored, is characterized in that when a processor executes the computer program, implementing the steps of a method comprising:

a ligand screening model construction method, which comprises:

obtaining a drug ligand training set, wherein the drug ligand training set includes a drug ligand chemical formula and one or more classification labels;

based on the ligand chemical formula, drawing a ligand graph network, in which atoms are nodes and chemical bonds are edges connecting the nodes;

using a random initialization vector to identify a weight vector of each node in the ligand graph network;

reconstructing each node of the ligand graph network according to a connection relationship of the ligand graph network to obtain a reconstruction graph network, and repeating the reconstruction step to obtain a reconstruction graph network comprising at least two layers;

performing deep learning on the ligand graph network and the reconstruction graph network comprising at least two layers, according to one or more classification labels; and

constructing the ligand screening model.

14 . The non-transitory computer readable storage medium of claim 13 , wherein using a random initialization vector to identify the weight vector of each node in the ligand graph network includes:

using at least one of a uniform distribution model and a normal distribution model to select a value of a (0,1) interval to generate the initialization vector; and

assigning the initialization vector to the node as the weight vector.

15 . The non-transitory computer readable storage medium of claim 13 , wherein reconstructing each node of the ligand graph network according to the connection relationship of the ligand graph network to obtain a reconstruction network includes:

obtaining the weight vector of each node and an associated weight vector of an associated node connected with the node according to the connection relationship of the ligand graph network;

integrating the weight vector and the associated weight vector together to obtain a reconstruction weight vector of the node; and

generating the reconstruction network according to the reconstruction weight vector.

16 . The non-transitory computer readable storage medium of claim 13 , wherein the ligand graph network and the reconstruction graph network comprising at least two layers are deeply learned according to the classification labels; and further comprising:

constructing a ligand screening model, which includes:

respectively extracting features of the ligand graph network and the reconstruction graph network comprising at least two layers, to obtain hierarchical network features;

according to the classification label, adopting a deep learning network to learn hierarchical network characteristics; and

constructing the ligand screening model.

17 . The non-transitory computer readable storage medium of claim 16 , wherein extracting features is performed on the ligand graph network and the reconstruction graph network comprising at least two layers respectively to obtain hierarchical network features, including:

performing feature extraction on the ligand graph network and the reconstruction graph network comprising at least two layers respectively through a function to obtain function vector features; and

normalizing the function vector features in probability space to obtain hierarchical network features.