Research On Methods Of Inferring Gene Regulatory Network Based On Gene Expression Profile

Gene regulatory network(GRN)is the important mechanism to maintain life process,control biochemical reaction and regulate compound level,which plays an important role in various organisms and systems.Reconstructing GRN can help us to understand the molecular mechanism of organisms,reveal essential rules of many biological processes and reactions.This paper classifies the existing GRN inference methods from model-based,information theory-based,and machine learning-based approaches,respectively,taking the classical,representative,and state-of-the-art methods in each category as examples to analyze the core ideas,general steps,characteristics,and so on.This paper comprehensively evaluates the performance of GRN inference algorithms through experiments on various datasets.This study briefly introduces and summarizes the existing methods of constructing GRN based on gene expression profile,helps researchers quickly understand the overall flow and latest developments of computational means in this field.Through a series of experimental evaluations,it summarizes the advantages and limitations of each method,avoids repetitive errors for subsequent researches,and puts forward feasible improvement directions.Additionally,this paper recommends the most appropriate computational methods for specific application scenarios and network types,so as to provide reference for biologists and medical scientists.Different from bulk expression data,single-cell transcriptomic data embody cellto-cell variance and diverse biological information,such as tissue characteristics,cell types,etc.Inferring GRN based on such data offers unprecedented advantages for making a profound study of cell phenotypes,revealing gene functions and exploring potential interactions.This paper proposes a hybrid deep learning framework for GRN inference from single-cell transcriptomic data,DGRNS,which encodes the raw data and fuses recurrent neural network(RNN)and convolutional neural network(CNN)to train a model capable of distinguishing related gene pairs from unrelated ones.Through a series of experiments on mouse and human datasets,it is proved that the accuracy of DGRNS is significantly better than other methods,and the generalization is strong.DGRNS solves the high sparsity problem of single cell data by means of sliding window,which provides a heuristic encoding way for the connection between biological data derived from experiments and artificial intelligence algorithms.In addition to identify the well-known regulatory relationships,DGRNS is able to infer a series of potential gene interactions with functional consistency,which will be the focus of verification by means of biological experiments.