Uncovering intratumor heterogeneity with Bayesian nonparametrics

Carcinogenesis is an evolutionary process in which advantageous mutations accumulate over time and cells harboring these mutations give rise to new clones. The nature of this progressing disease makes time course analysis very difficult and therefore the composition of these clones is generally unknown. We will discuss methods designed to identify intratumoral cancer clones with different sequencing paradigms, namely bulk and single-cell sequencing. A common foundation of the methods is Bayeisan nonparametrics, by which the number of clones can be inferred in a principled framework.

Ke works on statistical methods for quantifying breast cancer mircoenvironment from tumor images. His goal is to build probabilistic decision making machines to predict clinical outcomes (survival, relapse, etc), based on empirical features extracted from both tumor images and the associated molecular profiles (gene expression, copy number, etc).