The application of pangenomics and machine learning in genomic selection in plants

Abstract

Genomic selection approaches have increased the speed of plant breeding, leading to growing crop yields over the last decade. However, climate change is impacting current and future yields, resulting in the need to further accelerate breeding efforts to cope with these changing conditions. Here we present approaches to accelerate plant breeding and incorporate nonadditive effects in genomic selection by applying state-of-the-art machine learning approaches. These approaches are made more powerful by the inclusion of pangenomes, which represent the entire genome content of a species. Understanding the strengths and limitations of machine learning methods, compared with more traditional genomic selection efforts, is paramount to the successful application of these methods in crop breeding. We describe examples of genomic selection and pangenome-based approaches in crop breeding, discuss machine learning-specific challenges, and highlight the potential for the application of machine learning in genomic selection. We believe that careful implementation of machine learning approaches will support crop improvement to help counter the adverse outcomes of climate change on crop production.

Pangenomes as a Resource to Accelerate Breeding of Under-Utilised Crop Species

“…Machine Learning (ML) is a computational technology used to predict outcomes for specific problems based upon previous data. In bioinformatics, ML is becoming increasingly applied and optimised for crop-related advances in genomics and phenomics [ 118 , 130 , 131 , 132 ]. A recent study used random forest classification in conjunction with linkage disequilibrium mapping to identify pangenome PAV tags in domesticated barley with 83.6% accuracy, and in wild barley with 88.6% accuracy [ 133 ].…”

“…Furthermore, advancements in sequencing technologies will likely see pangenomes constructed with long-read DNA sequencing methods and chromosome-scale assemblies overtake single reference genomes for use in plant breeding research. The implementation of these pangenome assemblies in graph-based pangenomes and improvements in the accuracy of assembly and annotation tools will allow for more detailed analyses of the genetic constitution of under-utilised crops, and more efficient improvement of traits [ 88 , 92 , 131 , 156 ]. With pangenomes, existing genomic data and ML tools informing genetic breeding and gene editing, some of these climate-resilient and nutritious under-utilised crops show the potential to become alternative food sources or safety nets to major crops, supporting future increased agriculture system diversity and food security.…”

Understanding R Gene Evolution in Brassica

“…With more sequenced Brassica genomes available, the pangenome approach has become more attractive, considering the rich evolutionary information that can be deciphered at the genome and gene levels. It is currently feasible to use pangenomes with the implementation of state-of-the-art machine learning, not only to discover all the genes present in different Brassica species and explore the genome changes that happen in each of these polyploid species during the evolutionary pathway, but also to use this wealth of information in genomic selection, which offers a very promising future for Brassica crop breeding [144]. By developing pangenomes of B. napus and its progenitors, B. oleracea and B. rapa, it was revealed that defence-and stress-related genes are common dispensable genes and that gene loss events observed in specific Brassica plants are attributed to various evolutionary mechanisms [48].…”

Genomic prediction of agronomic traits in sugarcane using machine learning and quantitative genetics

“…( 2021) s ho wed t h at t h e accu r acy of models that considered nonadditive genetic effects, albeit only for some traits, was higher than that of models that considered additive effects alone. Islam et al (2021) demonstrated that the use of machine learning methods is an effective way to model nonadditive genetic effects (Bayer et al 2021). Typically, machine learning methods possess excellent potential to incorporate nonadditive effects and may be useful for genomic selection in sugarcane.…”