Prioritization of candidate genes in QTL regions associated with bioenergy-related traits in sorghum (Sorghum bicolor) using a machine learning algorithm

Abstract

Gene prioritization pipelines are designed to rank positional candidate genes (CG) within quantitative trait loci (QTL) and reduce the list of CG that is selected for further in-depth functional analysis. We have designed an integrated approach to prioritize CG in sorghum (Sorghum bicolor) combining the use of high-resolution QTL mapping, a machine learning algorithm, sequence analysis of the parental genomes and CG expression profiling. First, we re-mapped QTL associated with 20 different bioenergy-related traits in a recombinant inbred line (RIL) population from a cross between grain (M71) and sweet sorghum (SS79), genotyped using an Affymetrix 90K sorghum single nucleotide polymorphism (SNP) array. Thirty-eight QTL for 16 traits were identified using composite interval mapping; reference genome coordinates were determined for each QTL confidence interval and lists of positional CG generated. Positional CG lists were ranked using a machine learning algorithm, QTG-Finder2. Genomes of the RIL parental lines were re-sequenced in an Illumina NovaSeq 6000 (S4 flow cell, 300 cycles, PE150). Sequencing reads were aligned to the sorghum reference genome, BTx623, and SNPs were called for the parental genotypes. SNP effects on parental allele function were assessed using SNPeff. We also evaluated the tissue-specificity of each of the top 20% CG ranked by QTG-Finder2. Lastly, we generated a prioritized list of positional CG for each of the 38 QTL based on QTG-Finder2 rank, SNP presence/effect between parental alleles and expression profile. Taken together, these results bring us a step closer to finding the causal genes behind these set of bioenergy-associated traits.