Computational and functional dissection of drug targets in melanoma


Levi Garraway and Dana Pe’er




This project combines computational modeling with functional genomics to advance cancer therapy, using melanoma as an experimental model. We will use our Bayesian-network framework to integrate heterogeneous data types, including genotype, gene expression changes caused following perturbations, and drug resistance. This will let us: (1) Identify genetic alterations in tumors that drive proliferation and drug resistance; (2) Model how these alterations perturb normal cell growth and survival; (3) Understand the adaptive and feedback mechanisms that reduce drug efficacy, even with target inhibition; and (4) Identify additional target pathways for combinatorial drug treatment.


To provide an integrated view of tumor-specific genetic alterations, we scaled our Bayesian-network-based genetics-to-genomics framework from yeast to mammalian cells, and to cancer in particular. Specifically, we developed the CONEXIC algorithm to integrate copy-number and gene expression data to identify tumor-specific “driver” mutations, as well as the cellular processes they alter.

Using data from melanoma patients, we inferred a global map of 70 genes that are causally related to melanoma, and of the core processes their mutations alter. CONEXIC identified many known driver mutations of cancer, correctly coupling them with their known targets. To our surprise, several genes that were previously unrecognized as drivers of cancer are known to be involved in endosomal processes. We validated the computationally predicted role of these genes in melanoma by silencing them using short-hairpin RNA in short-term tumor-derived cultures. The results suggest that protein trafficking through endosomes may play an important role in this malignancy.

In another line of research, we used our computational framework to study the genetic and signaling networks that activate Erk in melanoma. Most melanomas are characterized by point mutations that activate the BRAF and NRAS oncogenes, suggesting that MAP kinase pathway inhibition may offer an appealing targeted therapeutic avenue in this malignancy, and about half of melanomas have a valine-to-glutamate mutation in codon 600 of the BRAF gene. BRAF V600E tumors resist drugs that inhibit Mek and thus activation of Erk. To study the Raf-Mek-Erk pathway in melanoma and to better understand this resistance, we studied several different genetically-defined tumors, measuring the time evolution of their expression with microarrays. We are analyzing this data by incorporating time variation into the computational tools. The results may let us computationally identify new druggable targets, include combinations of drugs that target different parts of the network.

Project Publications

Akavia UD, Litvin O, Kim J, Sanchez-Garcia F, Kotliar D, Causton HC, Pochanard P, Mozes E, Garraway LA, Pe'er D.  An integrated approach to uncover drivers of cancer. Cell. 2010 Dec 10;143(6):1005-17.

Sanchez-Garcia F, Akavia UD, Mozes E, Pe'er D. JISTIC: identification of significant targets in cancer. BMC Bioinformatics. 2010 Apr 14;11(1):189.