Connecting genotype to phenotype with machine learning and explainable AI

Translating genotypes to phenotypes is challenging because the genetic mechanisms underlying trait variation are complex. Nonetheless, recent progress in genomic prediction enables the use of genomic variation information to predict complex traits. Beyond genomic variation, additional omics data, such as transcriptome, methylome, and metabolomes, from a diversity panel of a species with distinct genetic background are also available. In this seminar, we will discuss opportunities and challenges in integrating these multi-omics datasets with machine learning to build models capable of predicting complex traits. We will also discuss the use of explainable AI methods to interpret the machine learning model and demonstrate how the results of interpretation allow a better understanding of the mechanistic bases, i.e., important genetic variants and their interactions, underlying traits of interests.

講者: Professor Shin-Han Shiu (Department of Plant Biology, Department of Comp. Math. Sci., & Engr. Michigan State University)

主持人: Dr. Ming-Jung Liu (劉明容博士)

時間:

地點:(中央研究院南部院區跨領堿研究大樓(I)122會議 室)