Comparison of actionable events detected in cancer genomes by whole-genome sequencing, in silico whole-exome and mutation panels

Abstract

Background: Next-generation sequencing is used in cancer research to identify somatic and germline mutations, which can predict sensitivity or resistance to therapies, and may be a useful tool to reveal drug repurposing opportunities between tumour types. Multigene panels are used in clinical practice for detecting targetable mutations. However, the value of clinical whole-exome sequencing (WES) and whole-genome sequencing (WGS) for cancer care is less defined, specifically as the majority of variants found using these technologies are of uncertain significance.

Patients and methods: We used the Cancer Genome Interpreter and WGS in 726 tumours spanning 10 cancer types to identify drug repurposing opportunities. We compare the ability of WGS to detect actionable variants, tumour mutation burden (TMB) and microsatellite instability (MSI) by using in silico down-sampled data to mimic WES, a comprehensive sequencing panel and a hotspot mutation panel.

Results: We reveal drug repurposing opportunities as numerous biomarkers are shared across many solid tumour types. Comprehensive panels identify the majority of approved actionable mutations, with WGS detecting more candidate actionable mutations for biomarkers currently in clinical trials. Moreover, estimated values for TMB and MSI vary when calculated from WGS, WES and panel data, and are dependent on whether all mutations or only non-synonymous mutations were used. Our results suggest that TMB and MSI thresholds should not only be tumour-dependent, but also be sequencing platform-dependent.

Conclusions: There is a large opportunity to repurpose cancer drugs, and these data suggest that comprehensive sequencing is an invaluable source of information to guide clinical decisions by facilitating precision medicine and may provide a wealth of information for future studies. Furthermore, the sequencing and analysis approach used to estimate TMB may have clinical implications if a hard threshold is used to indicate which patients may respond to immunotherapy.

Keywords: actionable mutations; cancer genomics; clinical genomics; microsatellite instability; precision oncology; tumour mutation burden (TMB); whole-genome sequencing.

Figure 1

Overview of drug–biomarker pairs per tumour type within the Cancer Biomarker Database used by the Cancer Genome Interpreter. (A) Bars which are to the right of the vertical line represent drugs which are either approved or in clinical trials, while bars to the left of the vertical line represent drugs which are either in pre-clinical or case report stages. n refers to the number of biomarker–drug combinations for each specified tumour type. (B) Number of biomarkers for which there are cancer type-specific FDA-approved drug allocations for and non-cancer type-specific. (C) Number of biomarkers for which there are cancer type-specific clinical trial drugs for and non-cancer type-specific. FDA, Food and Drug Administration; NCCN, National Comprehensive Cancer Network.

Figure 2

Repurposing potential of datasets analysed. Percentage of patients with cancer-specific and non-cancer-specific Food and Drug Administration (FDA)-approved (A) sensitive and (C) resistance biomarkers. Percentage of patients with (B) sensitive and (D) resistance biomarkers for drugs in clinical trials. Bars to the left of x = 0 indicate percentage of patients who can be prescribed an on-label drug, bars to the right of x = 0 indicate percentage of patients who could be prescribed an off-label drug. Off-label prescriptions are additive to the on-label prescriptions.

Figure 3

Comparison of sequencing platforms for the detection of actionable variants in cancer datasets analysed. Percentage of patients identified by the Cancer Genome Interpreter (CGI) as having (A) actionable variants, (B) actionable variants conferring drug sensitivity and (C) variants conferring drug resistance, stratified by sequencing platform. WES, CPanel and Panel represent in silico down-sampled regions of the exome capture kit, comprehensive panel and hotspot mutation panel kit. Solid diamonds joined by solid lines represent percentage of patients with variants for approved drugs only, and open diamonds with dashed lines represent percentage of patients with variants for drugs which are in clinical trials. Drug allocations used are non-cancer-specific (off-label). CN, copy number; CPanel, comprehensive panel; Panel, hotspot panel; WES, whole-exome sequencing; WGS, whole-genome sequencing.

Figure 4

Estimation of TMB by WGS, in silico WES, comprehensive and hotspot panels for 10 tumour types. (A) TMB calculated using all mutations. Values along the x-axis represent the TMB estimated using WGS data, and values along the y-axis represent the TMB estimated using in silico WES. (B) TMB correlation between in silico WES (all mutations) and in silico WES (non-synonymous mutations only). (C) TMB correlation between WGS and comprehensive panel (all mutations). (D) TMB correlation between WES and comprehensive panel (non-synonymous mutations only). (E) TMB correlation between comprehensive panel (non-synonymous mutations only) and comprehensive panel (all mutations). CPanel, comprehensive panel; NS, non-synonymous; TMB, tumour mutation burden; WES, whole-exome sequencing; WGS, whole-genome sequencing.

Supplementary Figure S1

Supplementary Figure S1: The number of actionable biomarkers in a variety of cancer types. (A) The number of biomarkers annotated in the cancer biomarker database that are present in and shared between the most tumour types. (B) The number of shared FDA- and NCCN-approved drugs between these solid tumours. GEJ adenocarcinoma, gastro-oesophageal junction adenocarcinoma.