Heterogeneity everywhere! The lists of clonal and sub-clonal aberrations found here and there in many tumors get longer and longer. But is this whole heterogeneity business actually useful for anything?
Actually it is, as we show in a paper that just came out in PLoS Medicine: Spatial and temporal heterogeneity in high-grade serous ovarian cancer: a phylogenetic reconstruction. And as a free bonus it comes with an editorial by Andy Beck from Harvard, who says “open access to large scale datasets is needed to translate knowledge of cancer heterogeneity into better patient outcomes” — right he is!
Phylogenetic analysis predicts patient survival
The figure below gives an overview of our study. We analyzed multiple copy-number profiles from initially 17 patients (some drop out for various reasons and we have 14 in the end). As a first step we computed pairwise distances between the copy-number profiles (more on this below) which gave us the square red distance matrix depicted below and opened up two lines of inquiry:
First of all, the distances can be used to build patient-specific evolutionary trees, which tell you something about the order of events in tumor evolution. But, maybe even more importantly, the distances also allowed us to compute summary measures of tumor heterogeneity, which could be linked to patient survival. And, yes indeed, the more heterogeneous the tumor, the worse the survival. (In 14 patients! Not really a definitive statement, I know.)
Call the MEDICC!
The key computational component of our study is MEDICC, which was published end of last year and computes the red matrix in the figure above. The code is available on bitbucket and has already found its first users outside my group. For example it has been used to great effect in the recent Big Bang paper.
What is MEDICC? Here a few lines from the paper abstract:
MEDICC [is] a method for phylogenetic reconstruction and heterogeneity quantification based on a Minimum Event Distance for Intra-tumour Copy-number Comparisons.
Using a transducer-based pairwise comparison function, we determine optimal phasing of major and minor alleles, as well as evolutionary distances between samples, and are able to reconstruct ancestral genomes.
But, hey, if you really want to know what we did, why not learn it directly from the master? Roland is explaining everything you’d ever want to know about MEDICC in this phyloseminar hosted by Erick Matsen from FHCRC.
Readers of this post also enjoy reading the original paper and citing it.