Talking Plant Science: David Kainer

The ARC Centre of Excellence for Plant Success in Nature and Agriculture is proud to bring you the next seminar in our Talking Plant Science series presented by Dr David Kainer.

Digging deeper into the GWAS signal with a little machine learning along the way

Genome Wide Association Studies, or GWAS, have become a standard tool for the discovery of the genetic basis of complex traits. For over a decade, results from GWAS have been used to guide experimentation, marker assisted selection and genetic engineering efforts. But for complex traits where we don’t have huge sample numbers (as with most plant studies!), GWAS outcomes can be very limited by multiple testing correction. Only loci that make it below the magic p-value threshold are deemed interesting. These loci often explain only a small fraction of the trait’s heritability, yet we know intuitively that many causal loci sit just ‘out of reach’. Here I will relate our efforts to relax those thresholds with the goal of reliably obtaining more of the trait genetic architecture. To deal with the peril of increasing false positives, multi-omic data sources such as gene expression and metabolic pathways can be fused into multiplex networks upon which network propagation algorithms tease apart the false positives from the true positives. I will demonstrate the process with examples in Arabidopsis and other species.

About the speaker: David is currently a Staff Scientist in the Computational Predictive Biology group at Oak Ridge National Laboratory, Tennessee USA. He specializes in analysing the genomic basis of complex traits to guide crop improvement through breeding and engineering. David received his PhD from the Australian National University’s Research School of Biology in 2017 after an earlier career in computer engineering and mobile game development. David focuses on biological network analysis, where multiple forms of ‘omic data are rendered as network layers and the combined (or Multiplexed) network is jointly traversed by machine learning algorithms such as Random Walks. This provides a platform for knowledge synthesis and discovery from highly complex, interconnected, heterogeneous data — a 21st century solution for a 21st century challenge.