Maize Carotenoid Gene Locus Mining Based on Conditional Gaussian Bayesian Network

Abstract

How to mine the gene locus for maize carotenoid components is an important research problem in biology study. Along with the rapid development of high-throughput biotechnologies, we have produced a large number of maize multi-omics data, including genome, transcriptome, metabolome, phenotype, etc. How to conjointly analyze these continuous and discrete data, and thus to mine the genetic loci that control the maize carotenoid components have a very important biological significance. In this work, we use the conditional Gaussian Bayesian network learning method to construct the network of maize gene, SNP locus and carotenoid components, aim to get the possible significant loci about four reported genes for the carotenoid component traits. The method is validated using the multi-omics data of maize global germplasm collection with 368 elite inbred lines. Four algorithms are used to do the comparison, and experiment results show the method can mine the effective locus for the phenotype traits. It is concluded that the conditional Gaussian Bayesian network learning method is an effective way of analyzing multi-omics data conjointly, mining the possible gene locus for maize carotenoid component traits, and thus to provide genetic resources and useful information for molecular breeding of maize. INDEX TERMS Gene locus mining, maize, conditional Gaussian Bayesian network, carotenoid components.

Artificial Intelligence in Food Safety: A Decade Review and Bibliometric Analysis

“…For instance, Zhao et al (2020) [ 135 ] chose the most optimal hyperparameters and kernel function for the SVM model to explore the genomic-based prediction performance in pigs and maize. To gain the possible significant loci, Liu et al (2020) constructed the maize gene, SNP locus, and carotenoid components network using the conditional Gaussian Bayesian network learning method [ 120 ]. Lv et al (2022) proposed minmax concave penalty (MCP) regularization for sparse deep neural networks (DNN-MCP), which can provide the optimal sparse structure for DNN and then greatly improve their ability to predict genomes, especially for the genomes of three quantitative traits [ 187 ].…”

Healthcare Cost Prediction Based on Hybrid Machine Learning Algorithms

“…In addition, researchers have used CGBN in other fields of research; for example, Hu [20] used CGBN to process seismic data with a mixture of discrete and continuous variables, then applied CGBN to the prediction of earthquakes in Canterbury from 2010 to 2011; the experimental results showed that the prediction performance of CGBN was better than that of algorithms such as neural networks and support vector machine. Liu in [21] used CGBN to mine gene loci for carotenoid components of maize, finding that CGBN exhibited better performance than other algorithms in the experiment. In this paper, CGBN is used to learn isolated feature variables (variables that do not affect the target variable) from healthcare expenditure data, then regression algorithms are used to learn the data with the isolated variables removed.…”