Comparative analysis of tissue-specific genes in maize based on machine learning models - CNN performs technically best, LightGBM performs biologically soundest

Aug 4, 2026

Upload by Logic

Comparative analysis evaluates how linear models (Limma), machine learning (LightGBM), and deep learning (CNN) identify tissue-specific genes from large-scale maize multi-tissue RNA-seq datasets. An expression matrix is processed using information gain and SHAP-based strategies, then validated by k-means clustering complementarity (V-measure), GO analysis, and literature retrieval. Results show CNN achieves higher technical performance, LightGBM highlights key transcription factors, and combined gene sets yield 78 biologically significant core tissue-specific genes.

LanguageEnglish

CategoryResearch & Report

FormatPDF

Pages15

Reading Time38 min

AccessFree to download

Introduction

Methods

Results

Discussion

TYPE Original Research PUBLISHED 09 May 2023 DOI 10.3389/fgene.2023.1190887 OPEN ACCESS EDITED BY Hao Cheng, University of California, Davis, United States REVIEWED BY Xiujun Zhang, Wuhan Botanical Garden (CAS), China Indika Kahanda, University of North Florida, United States *CORRESPONDENCE Zijie Wang, Comparative analysis of tissue-speciﬁc genes in maize based on machine learning models: CNN performs technically best, LightGBM performs biologically soundest [wangzj55@mail2.sysu.edu.cn](wangzj55@mail2.sysu.edu.cn) †These authors have contributed equally to this work and share last authorship RECEIVED 21 March 2023 ACCEPTED 17 April 2023 PUBLISHED 09 May 2023 CITATION Wang Z, Zhu Y, Liu Z, Li H, Tang X and Jiang Y (2023), Comparative analysis of tissue-speciﬁc genes in maize based on machine learning models: CNN performs technically best, LightGBM performs biologically soundest. Front. Genet. 14:1190887 . doi: 10.3389/fgene.2023.1190887 COPYRIGHT © 2023 Wang, Zhu, Liu, Li, Tang and Jiang. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY) . The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms. Zijie Wang 1*, Yuzhi Zhu 1, Zhule Liu 1†, Hongfu Li 1†, Xinqiang Tang 2† and Yi Jiang 1† 1School of Agriculture, Sun Yat-sen University, Shenzhen, China, 2School of Intelligent Systems Engineering, Sun Yat-sen University, Shenzhen, China Introduction: With the advancement of RNA-seq technology and machine learning, training large-scale RNA-seq data from databases with machine learning models can generally identify genes with important regulatory roles that were previously missed by standard linear analytic methodologies. Finding tissue-speciﬁc genes could improve our comprehension of the relationship between tissues and genes. However, few machine learning models fortranscriptome data have been deployed and compared to identify tissuespeciﬁc genes, particularly for plants. Methods: In this study, an expression matrix was processed with linear models (Limma), machine learning models (LightGBM), and deep learning models (CNN) with information gain and the SHAP strategy based on 1,548 maize multi-tissue RNA-seq data obtained from a public database to identify tissue-speciﬁc genes. In terms of validation, V-measure values were computed based on k-means clustering of the gene sets to evaluate their technical complementarity. Furthermore, GO analysis and literature retrieval were used to validate the functions and research status of these genes. Results: Based on clustering validation, the convolutional neural network outperformed others with higher V-measure values as 0.647, indicating that its gene set could cover as many speciﬁc properties of various tissues as possible, whereas LightGBM discovered key transcription factors. The combination of three gene sets produced 78 core tissue-speciﬁc genes that had previously been shown in the literature to be biologically signiﬁcant. Discussion: Different tissue-speciﬁc gene sets were identiﬁed due to the distinct interpretation strategy for machine learning models and researchers may use multiple methodologies and strategies for tissue-speciﬁc gene sets based on their goals, types of data, and computational resources. This study provided comparative insight for large-scale data mining of transcriptome datasets, shedding light on resolving high dimensions and bias difﬁculties in bioinformatics data processing. Frontiers in Genetics 01 [frontiersin.org](frontiersin.org) Wang et al. 10.3389/fgene.2023.1190887 KEYWORDS tissue speciﬁc genes, maize, CNN, limma, SHAP (shapley additive explanation),

FAQs

Which models were compared for identifying tissue-specific genes in maize?

The study compares linear models (Limma), machine learning models (LightGBM), and deep learning models (CNN).

How were the identified gene sets validated?

Validation used k-means clustering complementarity with V-measure, GO analysis, and literature retrieval to confirm gene functions and research relevance.

What were the main findings about model performance and biological significance?

CNN showed the best technical performance via higher V-measure, LightGBM uncovered key transcription factors, and combining gene sets produced 78 core tissue-specific genes previously reported as biologically significant.

Aug 4, 2026

Upload by Logic