Using The RBF Neural Network To Predict And Analyze Gene Expression And The Structure Of Gene Regulatory Networks

In recent years, there has been a significant focus on exploring gene regulatory networks from data, with a strong emphasis on developing techniques for predicting time series and classifying gene regulatory networks (GRNs) using gene expression data. Recently, a variety of neural networks have been introduced to support different scientific fields, with the RBF network standing out for its impressive performance in deciphering gene interactions. In this study, we generated synthetic time series gene expression data from a collection of GRNs and utilized a dualattention RNN to forecast the temporal behavior of genes. Our findings demonstrate the high accuracy of predictions for GRNs with diverse architectures. Furthermore, we delved into the RBF attention mechanism and leveraged specialized network tools, such as various graphs, to show that its graph features can distinguish between the hierarchical structures of distinct GRNs. We observed that GRNs exhibit varying responses to noise introduced in predictions by the RBF network, yielding highly dependable outcomes. This advanced network type allows for precise analysis of results and trends, paving the way for enhanced prediction of time series and inference of GRNs from gene expression data. The findings from both simulated and actual datasets validate the method's success in uncovering regulatory networks through accurately modeling the temporal changes in gene expression patterns

Keywords:

RBF neural network , gene regulatory network , time series prediction , dual attention