MiRNA-disease association prediction method based on attention mechanism

Abstract

According to the miRNA-disease association prediction method based on the attention mechanism, unknown miRNA-disease association is searched on the basis of known association data, and a foundation is laid for disease diagnosis and medicine research and development. The method comprises the following steps: firstly, obtaining initial feature representation of miRNA and diseases by utilizing a Gaussian profile kernel similarity function; embedding and longitudinally splicing miRNA and disease characteristics in the sample to form a sequence with only two nodes as the input of the encoder, an encoder being composed of a sub-attention module and a feedforward neural network module, and adding a residual module behind each sub-module in order to ensure the effectiveness of a result; and finally, using the output of the encoder as the score associated with the input prediction of the MLP decoder; and determining an optimal parameter through specific loss function back propagation.

Description

technical field [0001] The invention relates to a feature extraction method, in particular to a graph node feature extraction method based on a self-attention mechanism. Background technique [0002] With the rapid development of intelligent computing models, this technology has achieved remarkable results in the field of miRNA-disease prediction. Since it is time-consuming and labor-intensive to discover potential associations between miRNA-diseases using biological experiments, more and more researchers use advanced intelligent computing models to predict potential miRNA-disease associations. The core of solving this problem is how to extract effective feature representations of miRNAs and diseases. At present, the more commonly used method is to use graph neural network to obtain the feature representation of nodes. The main models include graph convolutional neural network, graph autoencoder, etc. [0003] Considering that there are certain differences in the interacti...

AI Technical Summary

Problems solved by technology

[0003] Considering that there are certain differences in the interaction between miRNA-diseases, it is necessary to choose an appropriate method to distinguish this difference in order to conform to the theoretical basis of reality, but most of the current methods ignore this point, resulting in poor predictive performance of the method. Difference

Images