Grant: Project Grant

Both for scientific studies to understand and find personalized cures for various subtypes of various cancers, and for diagnosis of individual patients with specific cancer markers or subtypes, a critical step is to sample patient tumors via some kind of tissue subsampling procedure (e.g. biopsy), to estimate the properties of the entire tumor. However, tumors are highly heterogeneous and spatially patterned, making it difficult to get accurate estimation of a whole tumor's properties with regional biopsy samples, for the same fundamental statistical reasons as why, e.g. political polling can be inaccurate when polling is concentrated in specific geographic regions. Using a combination of statistical theory, big imaging datasets of human tumors, and practical clinical expertise, we will derive methods to optimally diagnose patients using the least invasive biopsy procedures, and to boost the efficiency at which key drivers of cancer can be newly discovered in studies using patient tissue samples.