Matt Ploenzke, of the Department of Biostatistics at Harvard University and candidate for a faculty position in the Division of Biostatistics and the Childhood Cancer Genomics Group, will present:
“Redefining Pharmacogenomic Cell Sensitivity Using Multi-level Models”
Abstract: Pharmacogenomic experiments allow for the systematic testing of compounds at varying dosage concentrations to study how genomic markers correlate with cell sensitivity to treatment. While such methods hold potential for developing effective personalized treatment regimes, the current state of the field lacks a rigorous statistical framework for evaluating dose response both in terms of labelling which cells are sensitive/resistant to a given compound as well as how all available experimental data may be combined into a single dataset for biomarker discovery and validation. To this end we formulate a mixture model to estimate the drug-specific thresholds for discretizing cells into sensitive versus resistant and for classifying compounds into effect types (broad effect versus targeted effect). We motivate two formulations: 1) Fit independently to a given experiment and used to assess interstudy agreement and 2) Fit to all available experiments and used for biomarker testing. Case studies are presented to highlight important model distinctions and assess the agreement between the oft-compared CCLE and GDSC datasets.
A social tea will be held at 9:30 a.m. in A434 Mayo. All are Welcome.