PGD: a machine learning-based photosynthetic-related gene detection approach

Area of Science:

• Genomics
• Plant Biology
• Bioinformatics

Background:

• Photosynthetic capacity is crucial for crop yield, controlled by photosynthesis-related genes.
• Maize has a significant number of genes with unknown functions, potentially including novel photosynthetic genes.
• Machine learning offers a promising avenue for identifying these uncharacterized genes.

Purpose of the Study:

• To develop and validate a machine learning-based approach for detecting photosynthesis-related genes in maize.
• To identify novel candidate genes from the "not assigned" functional category in maize.
• To provide a reliable and robust method for mining genes involved in specific biological processes.

Main Methods:

• Utilized an ensemble machine learning model with a majority voting scheme.
• Integrated RNA-seq data from multiple photosynthetic mutants to enhance prediction accuracy.
• Developed the Photosynthetic-related Gene Detection approach (PGD) for identifying candidate genes.

Main Results:

• The ensemble model demonstrated improved prediction accuracy compared to single models.
• Successfully predicted 716 photosynthesis-related genes from the "not assigned" maize gene category.
• Validated predictions using protein localization (TargetP) and expression trends in maize leaf sections, confirming reliability.

Conclusions:

• The study presents a novel machine learning-based strategy for discovering genes within specific functional categories.
• Identified a set of 716 candidate genes for further experimental investigation into their roles in photosynthesis.
• The PGD approach and its online implementation offer a valuable tool for plant biology research.