Leveraging Transcriptomics-Based Approaches to Enhance Genomic Prediction: Integrating SNP weights and gene-networks for Cotton Fibre Quality Improvement

Abstract

The cotton genome (Gossypium hirsutum) contains [~] 80K protein-coding genes, making precision breeding for complex traits a challenge. This study tested biology-informed approaches to improve genomic prediction (GP) accuracy for cotton fibre traits to help accelerate precision breeding of valuable traits. The studys foundational approach was the use of RNA-seq data from key time points during fibre development, namely fibre cells undergoing primary, transition, and secondary wall development. The test approaches included using a range of summary statistics from RNA-seq analysis such as gene Differential Expression (DE). The three test approaches included DE genes overall, target pairwise DE lists informed by gene functional annotation, and finally, gene-network-clusters created based on Partial Correlation and Information Theory (PCIT) as the prior information in Bayesian GP models. The most promising improvements in GP accuracy were at the level of [~] 5% increase by using PCIT-based gene-network clusters as the prior knowledge network neighbours of key genes, and for the traits of cotton fibre Elongation and Strength. These results indicate that there is scope to help improve precision breeding of target traits by incorporating biology-based inference into GP models, and points to specific approaches to achieve this. KEY MESSAGEImproved Genomic Prediction accuracy of cotton fibre quality traits for helping accelerate precision breeding can be achieved by using biology-based prior knowledge and gene-network clusters.