CN-122024822-A - Method for predicting interaction between silencer and promoter and related equipment

Abstract

The application relates to the technical field of biological information, and provides a silencer and promoter interaction prediction method and related equipment, wherein the method comprises the steps of obtaining structural data of a plurality of chromatin ring structures; the method comprises the steps of respectively predicting a chromatin ring structure by utilizing an interaction prediction model based on structural data of the chromatin ring structure to obtain an interaction prediction result between a silencer and a promoter in the chromatin ring structure, constructing a loss function according to all the interaction prediction results, training the interaction prediction model by utilizing the loss function to obtain a trained interaction prediction model, and predicting the chromatin ring structure to be predicted by utilizing the trained interaction prediction model to obtain a final interaction prediction result between the silencer and the promoter in the chromatin ring structure to be predicted. The method can improve the accuracy of the interaction prediction of the silencer and the promoter.

Inventors

• LI MIN
• WANG ZIRUI
• LUO HANYU

Assignees

• 中南大学

Dates

Publication Date

20260512

Application Date

20260203

Claims (10)

1. 1. A method for predicting the interaction of a silencer with a promoter, comprising: obtaining structural data of a plurality of chromatin loop structures, wherein the structural data comprises promoter positions and silencer positions in the corresponding chromatin loop structures; Respectively predicting structural data of the chromatin ring structures by utilizing an interaction prediction model according to each chromatin ring structure to obtain an interaction prediction result between a silencer and a promoter in the chromatin ring structure; constructing a loss function according to all interaction prediction results, and training the interaction prediction model by using the loss function to obtain a trained interaction prediction model, wherein the loss function is used for describing the accuracy of all interaction prediction results; and predicting the chromatin ring structure to be predicted by using the trained interaction prediction model to obtain a final interaction prediction result between the silencer and the promoter in the chromatin ring structure to be predicted.
2. 2. The method according to claim 1, wherein the predicting, based on the structural data of the chromatin loop structure, using an interaction prediction model, results in a predicted interaction between a silencer and a promoter in the chromatin loop structure, comprising: obtaining a promoter sequence from the chromatin loop structure according to the promoter position in the structural data; Obtaining a silencer sequence from the chromatin loop structure according to silencer positions in the structural data; and predicting based on the promoter sequence and the silencer sequence by using the interaction prediction model to obtain an interaction prediction result between the silencer and the promoter in the chromatin loop structure.
3. 3. The method of claim 2, wherein predicting based on the promoter sequence and the silencer sequence using the interaction prediction model results in a predicted interaction between silencer and promoter in the chromatin loop structure, comprising: converting the promoter sequence by utilizing the interaction prediction model to obtain a promoter embedding matrix, and converting the silencer sequence to obtain a silencer embedding matrix; extracting features of the promoter embedding matrix to obtain a promoter feature vector, and extracting features of the silencer embedding matrix to obtain a silencer feature vector; Splicing the promoter characteristic vector and the silencer characteristic vector to obtain splicing characteristics; And predicting based on the splicing characteristics to obtain an interaction prediction result between the silencer and the promoter in the chromatin loop structure.
4. 4. The method of interaction prediction according to claim 3, wherein the feature extraction of the promoter embedding matrix to obtain a promoter feature vector comprises: Extracting spatial features of the promoter embedded matrix to obtain a pooling feature matrix; And extracting time sequence characteristics of the pooling characteristic matrix to obtain a promoter characteristic vector.
5. 5. The method of interaction prediction according to claim 4, wherein the performing spatial feature extraction on the promoter embedding matrix to obtain a pooled feature matrix comprises: And carrying out continuous two-layer convolution operation on the promoter embedded matrix to obtain a convolution feature matrix, and carrying out maximum pooling operation on the convolution feature matrix to obtain a pooling feature matrix.
6. 6. The interaction prediction method according to claim 4, wherein the performing time sequence feature extraction on the pooled feature matrix to obtain a promoter feature vector comprises: By the formula: Calculate the first Final characteristics of individual time steps ; Wherein, the , For the last time step, when At the time of the first Final characteristics of individual time steps In order for the promoter to be a feature vector, In order to update the door(s), In order to reset the gate, In order to pool the feature matrix, Represent the first The hidden characteristics of the individual time steps, 、 、 As a matrix of weights, the weight matrix, 、 、 In order for the offset to be a function of, Represent the first Final characteristics of the individual time steps.
7. 7. The interaction prediction method according to claim 1, wherein the loss function is: Wherein, the In order to be a value of the loss function, Indicating the number of interaction prediction results, Represent the first The result of the prediction of the interaction is, Represent the first And the actual label corresponding to the interaction prediction result.
8. 8. A silencer and promoter interaction prediction device, comprising: the acquisition module is used for acquiring structural data of a plurality of chromatin loop structures, wherein the structural data comprises promoter positions and silencer positions in the corresponding chromatin loop structures; The first prediction module is used for predicting the structure data of the chromatin ring structures by utilizing an interaction prediction model according to each chromatin ring structure to obtain an interaction prediction result between a silencer and a promoter in the chromatin ring structures; The training module is used for constructing a loss function according to all interaction prediction results, and training the interaction prediction model by utilizing the loss function to obtain a trained interaction prediction model; And the second prediction module is used for predicting the chromatin ring structure to be predicted by using the trained interaction prediction model to obtain a final interaction prediction result between the silencer and the promoter in the chromatin ring structure to be predicted.
9. 9. A terminal device comprising a memory, a processor and a computer program stored in the memory and executable on the processor, characterized in that the processor implements the method for predicting the interaction of a silencer with a promoter according to any of claims 1 to 7 when executing the computer program.
10. 10. A computer readable storage medium storing a computer program, wherein the computer program when executed by a processor implements the method of predicting the interaction between a silencer and a promoter according to any one of claims 1 to 7.

Description

Method for predicting interaction between silencer and promoter and related equipment Technical Field The application relates to the technical field of biological information, in particular to a silencer and promoter interaction prediction method and related equipment. Background In genomics research, it is a significant challenge to identify functional elements in the genome, especially those that regulate the switches of gene expression (e.g., promoters, enhancers, and silencers). Wherein, the silencer is used as a key remote control element to inhibit gene expression by forming three-dimensional space interaction with a target promoter, and plays an irreplaceable role in cell fate determination, homeostasis maintenance and disease pathogenesis. The study of silencer-promoter interactions (SPIs, silencer-Promoter Interactions) was significantly delayed compared to the enhancer study, mainly due to the difficulty of experimental validation and the lack of high quality baseline data sets. While some methods exist in the prior art for identifying silencer candidate regions (e.g., based on DNase data or histone modification marker screening), most of these methods ignore critical three-dimensional chromatin structures that mediate silencer function. In addition, the currently mainstream calculation method is only focused on the identification of the silencer, or the integration of the DNA sequence characteristics and the three-dimensional chromatin structure cannot be simultaneously considered, so that the accuracy of the interaction prediction of the silencer and the promoter is low. Disclosure of Invention The embodiment of the application provides a method and related equipment for predicting the interaction between a silencer and a promoter, which can solve the problem of low accuracy of the interaction prediction between the silencer and the promoter. In a first aspect, embodiments of the present application provide a method for predicting the interaction between a silencer and a promoter, comprising: Obtaining structural data of a plurality of chromatin loop structures, wherein the structural data comprises promoter positions and silencer positions in the corresponding chromatin loop structures; Respectively predicting structural data of the chromatin loop structures by utilizing an interaction prediction model according to each chromatin loop structure to obtain an interaction prediction result between a silencer and a promoter in the chromatin loop structure; Constructing a loss function according to all interaction prediction results, and training the interaction prediction model by using the loss function to obtain a trained interaction prediction model; And predicting the chromatin ring structure to be predicted by using the trained interaction prediction model to obtain a final interaction prediction result between the silencer and the promoter in the chromatin ring structure to be predicted. Optionally, predicting based on structural data of the chromatin loop structure using an interaction prediction model to obtain an interaction prediction result between the silencer and the promoter in the chromatin loop structure, including: obtaining a promoter sequence from the chromatin loop structure according to the promoter position in the structural data; Obtaining a silencer sequence from the chromatin loop structure according to silencer positions in the structural data; and predicting based on the promoter sequence and the silencer sequence by using an interaction prediction model to obtain an interaction prediction result between the silencer and the promoter in the chromatin loop structure. Optionally, predicting based on the promoter sequence and the silencer sequence using an interaction prediction model to obtain a predicted result of the interaction between the silencer and the promoter in the chromatin loop structure, including: converting the promoter sequence by utilizing an interaction prediction model to obtain a promoter embedding matrix, and converting the silencer sequence to obtain a silencer embedding matrix; extracting features of the promoter embedding matrix to obtain a promoter feature vector, and extracting features of the silencer embedding matrix to obtain a silencer feature vector; Splicing the characteristic vector of the promoter and the characteristic vector of the silencer to obtain splicing characteristics; and predicting based on the splicing characteristics to obtain an interaction prediction result between the silencer and the promoter in the chromatin loop structure. Optionally, feature extraction is performed on the promoter embedding matrix to obtain a promoter feature vector, including: Extracting spatial features of the promoter embedded matrix to obtain a pooling feature matrix; And extracting time sequence features of the pooled feature matrix to obtain a promoter feature vector. Optionally, performing spatial feature extraction on the promoter embedded matrix