Articles
Identifying epistatic loci for fruit rot resistance in cranberry using machine learning
Article number
1357_23
Pages
157 – 162
Language
English
Abstract
Currently, most marker assisted selection (MAS) methodologies rely on markers developed from either quantitative trait loci (QTL) studies or genome wide association (GWAS) studies.
However, there is ongoing research demonstrating that many agriculturally important phenotypes are not conditioned by a single or few loci easily identified using QTL and GWAS methods, but often are the result of combinatorial interacting variants of multiple loci, e.g., epistatic loci.
Machine learning (ML) methods offer an alternative means of providing informative markers for genome selection and efficiently predicting interacting and epistatic markers.
We have developed methodology that utilizes ML and genotyping by sequencing to provide informative markers for MAS. Our method utilizes genotyping by sequencing data in common vcf format and a boosted regression trees algorithm in an easily obtained Python package.
In addition, we utilized ML methods to study the polygenic trait of fruit rot resistance in two populations of cranberry that were not well characterized using conventional QTL analysis software.
Four major epistatic loci on four different linkage groups were identified in these populations, with three loci contributing up to 23% of phenotypic variance in one population, and two loci contributing up to 27% of the variance in the other population.
SNPs at the loci identified facilitated the creation of SNP-based markers for genotyping and seedling screening.
These markers are now being validated in subsequent breeding cycles.
However, there is ongoing research demonstrating that many agriculturally important phenotypes are not conditioned by a single or few loci easily identified using QTL and GWAS methods, but often are the result of combinatorial interacting variants of multiple loci, e.g., epistatic loci.
Machine learning (ML) methods offer an alternative means of providing informative markers for genome selection and efficiently predicting interacting and epistatic markers.
We have developed methodology that utilizes ML and genotyping by sequencing to provide informative markers for MAS. Our method utilizes genotyping by sequencing data in common vcf format and a boosted regression trees algorithm in an easily obtained Python package.
In addition, we utilized ML methods to study the polygenic trait of fruit rot resistance in two populations of cranberry that were not well characterized using conventional QTL analysis software.
Four major epistatic loci on four different linkage groups were identified in these populations, with three loci contributing up to 23% of phenotypic variance in one population, and two loci contributing up to 27% of the variance in the other population.
SNPs at the loci identified facilitated the creation of SNP-based markers for genotyping and seedling screening.
These markers are now being validated in subsequent breeding cycles.
Publication
Authors
J. Kawash, J. Polashock, J. Johnson-Cicalese, N. Vorsa
Keywords
GWAS, QTL, GBS, machine learning, epistasis
Online Articles (56)