Decoding cancer heterogeneity: Using an information-theoretic approach to design patient-specific drug combinations

IntroductionWorld-wide attempts are invested, aiming to develop effective strategies for personalized cancer medicine. We study a proteomic dataset obtained from ~3500 patients of 11 types. We aim to explore the data space of... [ view full abstract ]

Introduction

World-wide attempts are invested, aiming to develop effective strategies for personalized cancer medicine. We study a proteomic dataset obtained from ~3500 patients of 11 types. We aim to explore the data space of cancer patients and find a way to accurately classify tumors, such that every single tumor can be mapped precisely and unambiguously according to the molecular aberrations that it harbors.

Methods

We study cancer from an information-theoretical point of view. Using information-theoretic surprisal analysis we identify in each tumor protein unbalanced networks in which a regular flow of biological information is disturbed leading to a deviation of the tissue from the balanced “homeostatic” state (Figure 1). This imbalance governs the survival and progression of the tumors.

 Results

We unraveled a surprisingly simple order that underlies the extreme complexity of tumor tissues, and demonstrated that only 17 unbalanced processes characterize this large and diverse collection of tumors. Each tumor is described by a specific subset of 1-4 processes out of 17. We show that the majority of tumor-specific sets, named barcodes, are extremely rare, and are shared by only 5 tumors or less, supporting a personalized, comprehensive study of tumors in order to design the optimal combination therapy for every patient. We suggest that inhibition of the entire set of tumor-specific networks should stop the disease and significantly decrease the chances for the development of drug resistance.

We experimentally validated our approach using 10 different cancer cell lines. Using surprisal analysis each cell line was assigned a set of unbalanced processes and a combined drug therapy (Figure 2). Figure 2 shows that therapies predicted by us worked efficiently even in triple negative breast cancer cell lines, for which targeted therapy is currently unavailable.

Discussion 

We present a novel approach to deal with the inter-tumor heterogeneity and to break down the high complexity of cancer systems into simple, easy to crack, barcodes (Figure 3). We sort the needles from the haystack, by identifying with high resolution patient-specific unbalanced processes and rewired signaling pathways. This deep understanding of tumor-specific imbalances should greatly advance the fields of cancer research and therapeutics.