1. MegaLMM: Mega-scale linear mixed models for genomic predictions with thousands of traits
- Author
-
Hao Cheng, Lorin Crawford, Daniel E. Runcie, and Jiayi Qu
- Subjects
0106 biological sciences ,Multivariate statistics ,Computer science ,Arabidopsis ,Method ,QH426-470 ,computer.software_genre ,01 natural sciences ,Models ,2.5 Research design and methodologies (aetiology) ,Aetiology ,Biology (General) ,Triticum ,0303 health sciences ,Genome ,food and beverages ,Genomics ,Biological Sciences ,Phenotype ,High-throughput phenotyping ,Genome, Plant ,Mixed model ,Linear mixed effect model ,Scale (ratio) ,Genotype ,Bioinformatics ,QH301-705.5 ,Biology ,Machine learning ,Mega ,Zea mays ,Generalized linear mixed model ,03 medical and health sciences ,Quantitative Trait ,Quantitative Trait, Heritable ,Genetic ,Information and Computing Sciences ,Genetics ,Leverage (statistics) ,Humans ,Plant breeding ,Heritable ,030304 developmental biology ,Genomic prediction ,Models, Genetic ,Multi-environment trial ,business.industry ,Human Genome ,Bayes Theorem ,Quantitative genetics ,Plant ,Multi-trait Linear Mixed Model ,Human genetics ,Genetic architecture ,Plant Breeding ,Gene-Environment Interaction ,Artificial intelligence ,business ,computer ,Software ,Environmental Sciences ,010606 plant biology & botany - Abstract
Plant breeding, like other fields in applied quantitative genetics, has embraced large-scale phenotype data as a way to rapidly and accurately create the next generations of crops. High-throughput phenotyping technologies and large-scale multi-environment breeding trials generate an unprecedented scale of data for breeders to use to make selections and crosses in breeding programs. However, the statistical foundation of multi-trait breeding is based on the multivariate linear mixed effect model, a tool notorious for its fragility when applied to more than a handful of traits. We present MegaLMM, a statistical framework and associated software package that facilitates flexible mixed model analyses on a virtually unlimited number of traits. Using three examples with real plant data, we show that MegaLMM can efficiently model the genetic architecture of thousands of traits at once while significantly improving genetic value prediction accuracy.
- Published
- 2021