Computational approaches for modelling heterogeneity across complex biological phenotypes

Abstract

Complex diseases such as cancer are typified by heterogeneity: both at the molecular level and in therapeutic response. This is seen clearly in the analysis of large molecular assays of gene and protein expression where findings are strongly dependent on the methodological approach used. Unsuper-vised analysis approaches are one way to measure and visualise biological heterogeneity in expres-sion data, especially for studies with large sample sizes or indeed integrative analysis of such large-scale datasets. However, it is a real challenge to assess how well such methods work given their high-dimensional nature and oftentimes a lack of a ground truth. In three independent projects, I've con-..