RiceSNP-ABST: a deep learning approach to identify abiotic stress-associated single nucleotide polymorphisms in rice

Area of Science:

• Genetics
• Bioinformatics
• Plant Science

Background:

• Abiotic stresses significantly impact rice production, necessitating the development of stress-resistant varieties.
• Accurate identification of single nucleotide polymorphisms (SNPs) associated with abiotic stress traits (ABST-SNPs) is vital for rice breeding.
• Current research faces challenges due to limited high-quality abiotic stress data in rice and the under-application of deep learning in this area.

Purpose of the Study:

• To develop an efficient computational model for predicting ABST-SNPs in rice.
• To address the scarcity of data and computational tools for identifying genetic variants related to rice abiotic stress tolerance.
• To facilitate the breeding of improved rice varieties with enhanced resistance to environmental challenges.

Main Methods:

• A novel negative sample construction strategy was employed to generate six training datasets.
• Four feature encoding methods based on DNA sequence fragments were proposed, followed by feature selection.
• Convolutional neural networks with residual connections were utilized for ABST-SNP prediction, alongside multi-granularity causal structure learning.

Main Results:

• The proposed RiceSNP-ABST model demonstrated superior performance compared to traditional machine learning and existing state-of-the-art methods.
• The model exhibited robust generalization capabilities on independent and cross-species datasets.
• Multi-granularity causal structure learning helped elucidate relationships among DNA structural features for effective identification of key genetic variants.

Conclusions:

• RiceSNP-ABST provides an effective deep learning-based approach for predicting ABST-SNPs in rice.
• The model's performance and generalization suggest its utility in accelerating rice improvement and breeding programs.
• The developed web-based tool (http://rice-snp-abst.aielab.cc) offers a valuable resource for researchers in rice genetics and breeding.