The combinatorial complexity of cancer precision medicine

Abstract

Precision medicine approaches have recently been developed that offer therapies targeting mainly single genetic alterations in malignant tumors. However, next generation sequencing studies have shown that tumors normally harbor multiple genetic alterations, which could explain the so far limited successes of personalized medicine, despite considerable benefits in certain cases. Combination therapies may contribute to a solution, but will pose a major challenge for clinical trials evaluating those therapies. As we discuss here, reasons include the low abundance of most of the relevant mutations and particularly the combinatorial complexity of possible combination therapies. Our report provides a systematic and quantitative account of the implications of combinatorial complexity for cancer precision medicine and clinical trial design. We also present an outlook on how systems biological approaches may be harnessed to contribute to a solution of the complexity challenge by predicting optimal combination therapies for individual patients and how clinical trial design may be adapted by combining and extending basket and umbrella design features.

Keywords: Clinical Trial Design; Combination Therapies; Personalized Therapy; Precision Medicine; Systems Medicine.

Figure 1. Combinatorial complexity of combination therapies in personalized medicine

A: Heatmap visualization of the number of possible combination therapies in dependence on the number of actionable mutations and number of combined drugs. As an example a combination therapy with 3 drugs selected out of a set of 40 compounds yields 9,880 possible combination therapies that would have to be evaluated clinically. B: Heatmap visualization of the number of patients that would have to be screened (i. e. whose tumors would have to be sequenced) to recruit 200 patients into a clinical trial evaluating a combination therapy in dependence on the number of actionable mutations targeted in the combination therapy and the frequency at which they occur in tumors.

Figure 2. A novel approach to clinical trials: combining and extending basket and umbrella trials

The classical approach recruiting patients according to a high-level diagnosis (e. g. lung adenocarcinoma “blue”) potentially refined by single markers (e. g. EML4-ALK-positive lung adenocarcinoma, “blue with magenta spot”) for statistical comparison between therapy groups (A) will fail with even a handful of druggable mutations under investigation (B) With an increasing number of actionable mutations and the need for combination therapies, even novel approaches such as basket or umbrella trials are incapable of addressing the arising combinatorial complexity. We therefore propose a concept draft for the development of a novel clinical trial design approach (C) that incorporates 1) a comprehensive functional analysis of the molecular tumor features that are 2) subsequently analyzed using bioinformatics and computational modeling of the (pathologically altered) network to identify target molecules. In combination with a drug library this knowledge is 3) used to propose optimal combination therapies for each patient who is then 4) recruited to the trial in which multiple different combination therapies are assessed. Such trials test the efficacy of the molecular analysis and therapy selection method and the employed drug library and therefore provide an implicit therapy validation.