G2PDeep: a web-based deep-learning framework for quantitative phenotype prediction and discovery of genomic markers

Area of Science:

• Genomics
• Bioinformatics
• Machine Learning

Background:

• Quantitative trait prediction is crucial for crop and animal breeding.
• Identifying genomic markers associated with phenotypes aids genetic studies.
• Existing tools may lack integrated deep learning frameworks for these tasks.

Purpose of the Study:

• To introduce G2PDeep, an open-access web server for quantitative phenotype prediction and genomic marker discovery.
• To provide researchers with a user-friendly platform for creating and training deep learning models.
• To leverage high-performance computing for efficient model development and deployment.

Main Methods:

• Utilizes zygosity or single nucleotide polymorphism (SNP) data as input.
• Employs a deep learning framework for prediction tasks.
• Offers an interactive web interface for model creation and training.
• Integrates soybean nested association mapping (SoyNAM) and public seed trait datasets.

Main Results:

• G2PDeep provides a comprehensive platform for end-to-end deep learning model development.
• The framework enables prediction of quantitative phenotypes and associated genomic markers.
• Benchmarking demonstrates competitive performance in quantitative phenotype prediction.
• Integrated datasets include soybean yield, height, moisture, oil, and protein content.

Conclusions:

• G2PDeep serves as a valuable resource for researchers in plant and animal genetics.
• The platform facilitates the discovery of genetic underpinnings of quantitative traits.
• G2PDeep promotes accessible application of deep learning in genomics research.