Topic Overview:
Genome scale technologies can measure thousands, or even millions, of molecules, but the individual measurements are often noisy readouts of the internal biological state. For many computational approaches, the ultimate goal is to infer the true biological state parameters from their effects on a much larger set of experimental readouts. Indeed, much of genomic data analysis can be viewed as a special case of dimensionality reduction. For biological data, the goal is to reduce the thousands or millions of genomic measurements to an interpretable representation of the underlying biological mechanisms.

Chikina and her group develop methods that incorporate intuition about biological processes into generic machine-learning models. In her lecture, Chikina will discuss her recent method, Pathway-level Information ExtractoR (PLIER), that produces pathway-level representations of gene expression data. This method infers cell-type proportions and pathway activities from any gene expression dataset, outperforming dedicated approaches. She will discuss how PLIER analysis can aid in interpreting complex genomic analyses like expression quantitative trait loci (eQTLs), as well as improve the concordance of phenotype association across different studies.

Continuing Medical Education (CME) credit: To receive CME credit for this lecture, text the code CASBES to 412-312-4424 within 24 hours of the session’s start. (Your mobile phone number first must be linked to your account at cce.upmc.com.) You can also email kbg12@pitt.edu to receive credit for this session.