Pan-cancer detection of driver genes at the single-patient resolution

Abstract

Identifying the complete repertoire of genes that drive cancer in individual patients is crucial for precision oncology. Established methods for driver detection focus mostly on genes that are recurrently altered across cohorts of cancer patients.However, mapping these genes back to patients leaves a sizeable fraction with few or no driver events, hindering our understanding of cancer mechanisms and limiting the choice of therapeutic interventions. Here we present sysSVM2, a tool based on machine learning that integrates somatic alteration data with systems-level gene properties to predict drivers in individual patients. We develop sysSVM2 for pancancer applicability, demonstrating robust performance on real and simulated cancer data. We benchmark its performance against other driver detection methods and show that sysSVM2 has a lower false positive rate and superior patient driver coverage. Applying sysSVM2 to 7,646 samples from 34 cancer types, we find that predicted drivers are often rare or patient-specific. However, they converge to disrupt well-known cancer-related processes including DNA repair, chromatin organisation and the cell cycle. sysSVM2 is a resource to enhance personalised predictions of cancer driver events with possible use in research and clinical settings. Code to implement sysSVM2 and the trained models in simulated cancer-agnostic data as well as in 34 cancer types are available at https://github.com/ciccalab/sysSVM2.

Comparative assessment of genes driving cancer and somatic evolution in noncancer tissues: an update of the NCG resource

“…This gap can be solved by complementing cohort-level approaches with methods that account for all types of alterations and predict drivers in individual samples, for example identifying their network deregulations [63][64][65] or applying machine learning to identify driver alterations 66 . Alternatively, we have shown that systems-level properties capture the main features of cancer drivers, justifying their use for patient-level driver detection 67,68 .…”

Pan-cancer detection of driver genes at the single-patient resolution

“…In this work, we developed a cancer-agnostic algorithm, sysSVM2, for identifying cancer driver in cancer individual patients [28]. By refining the machine learning approach upon which the original algorithm was built [18], we broadened its applicability to the pan-cancer range of malignancies represented in TCGA.…”

“…Based on these results, we chose the default settings for the cancer agnostic SVM classifier, which we named sysSVM2 [28]. By default, data are un-centred but scaled to have unit standard deviation.…”

“…In order to provide a comprehensive resource of trained models [28] and patient-level drivers, we sought to apply sysSVM2 to 7646 TCGA samples of 34 cancer types with at least one somatically damaged gene (Additional file 1: Supplementary Methods).…”

“…Platform-independent R code to implement sysSVM2, along with a README file and an example dataset, is available at https://github.com/ciccalab/ sysSVM2 [28]. The recommended settings as described in this manuscript are set as default values.…”