Related Experiment Video
Updated: Jul 29, 2025
10:08
Experimental Design for Laser Microdissection RNA-Seq: Lessons from an Analysis of Maize Leaf Development
Published on: March 5, 2017
9.6K
Comparative analysis of tissue-specific genes in maize based on machine learning models: CNN performs technically
Zijie Wang1, Yuzhi Zhu1, Zhule Liu1
1School of Agriculture, Sun Yat-sen University, Shenzhen, China.
Frontiers in Genetics
|May 25, 2023
Summary
Machine learning models, including convolutional neural networks (CNN), effectively identify tissue-specific genes in maize RNA-seq data. This approach surpasses linear methods, revealing core genes crucial for understanding plant tissue biology.
Area of Science:
- Bioinformatics
- Plant Genomics
- Machine Learning
Background:
- RNA sequencing (RNA-seq) and machine learning (ML) advance gene discovery.
- Identifying tissue-specific genes enhances understanding of gene-tissue relationships.
- Limited comparative studies exist for ML models in plant transcriptome analysis.
Purpose of the Study:
- To compare linear, ML, and deep learning models for identifying tissue-specific genes in maize.
- To evaluate the effectiveness of different models in analyzing large-scale RNA-seq data.
- To identify core tissue-specific genes with biological significance.
Main Methods:
- Analyzed 1,548 maize multi-tissue RNA-seq datasets using Limma, LightGBM, and CNN.
- Employed information gain and SHAP strategies for gene identification.
- Validated gene sets using k-means clustering (V-measure), GO analysis, and literature review.
Main Results:
- CNN achieved the highest V-measure (0.647), indicating superior coverage of tissue-specific properties.
- LightGBM identified key transcription factors.
- A combined approach yielded 78 core tissue-specific genes with documented biological significance.
Conclusions:
- Different ML models yield distinct tissue-specific gene sets due to varying interpretation strategies.
- Multiple methodologies can be employed for tissue-specific gene discovery based on research goals.
- This study offers comparative insights for high-dimensional transcriptome data mining in bioinformatics.
