BBSRC-NSF/BIO: TSLMM: Methodology, software, and applications for tree-sequence-based linear mixed modelling

This project addresses urgent computational and methodological challenges in modern genomics. It aims to develop new ways to analyze very large genomic datasets by combining advanced genetic models with detailed maps of how genomes have been inherited over time. The work aims to create fast, open-source software that can handle millions of genomes efficiently, test the methods on both simulated and real-world data, and provide training for researchers. The resulting tools could improve the prediction of important traits and thus will have benefits across many fields in biology, including agriculture (breeding efficiency) and medicine (genomic risk prediction). The goals proposed in this project will help researchers make better use of modern AI methods, and support advances in biotechnology and genetics research. This project sits at the intersection of quantitative genetics, statistical modelling and high-performance computing. It uses the tree sequence representation of Ancestral Recombination Graphs to implement the class of linear mixed models used across a range of quantitative genetics applications from genomic prediction (in breeding) to genome-wide association studies. Ancestral Recombination Graphs contain both time information (by estimation of mutation and common ancestor ages) and haplotype structuring and so provide a biologically descriptive means by which to formulate models. The project focuses on genomic prediction, where traditional reliance on large genetic relatedness matrices has hampered applications to modern datasets with millions of individuals. In particular, the goals are to (1) develop theory and algorithms for applying linear mixed models to tree-sequence encoded genome data; (2) create fast and scalable open-source software implementing these new approaches; (3) test the method using simulated data and large real datasets; and (4) provide training and tools for potential users of the methods. By combining tree sequences with linear mixed models, the resulting tools will allow scientists to work directly with large-scale genome datasets more efficiently, to learn more actionable information, and to better apply modern AI tools for biotechnology, thereby advancing NSF's priorities in these two fields. This award is made possible through the NSF-UKRI lead agency opportunity. This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria. NSF Award ID: 2624219 | Program: 01002627DB NSF RESEARCH & RELATED ACTIVIT | Principal Investigator: Peter Ralph | Institution: University of Oregon Eugene, EUGENE, OR | Award Amount: $553,321 View on NSF Award Search: https://www.nsf.gov/awardsearch/show-award/?AWD_ID=2624219 View on Research.gov: https://www.research.gov/awardapi-service/v1/awards/2624219.html