Analysis of Shared Variants between Cancer Biospecimens

Abstract

Purpose: Mutational data from multiple solid and liquid biospecimens of a single patient are often integrated to track cancer evolution. However, there is no accepted framework to resolve if individual samples from the same individual share variants due to common identity versus coincidence.

Experimental design: Utilizing 8,000 patient tumors from The Cancer Genome Atlas across 33 cancer types, we estimated the background rates of co-occurrence of mutations between discrete pairs of samples across cancers and by cancer type. We developed a mutational profile similarity (MPS) score that uses a large background database to produce confidence estimates that two tumors share a unique, related molecular profile. The MPS algorithm was applied to randomly paired tumor profiles, including patients who underwent repeat solid tumor biopsies sequenced with Memorial Sloan Kettering-IMPACT (n = 53,113). We also evaluated the MPS in sample pairs from single patients with multiple cancers (n = 2,012), as well as patients with plasma and solid tumor variant profiles (n = 884 patients).

Results: In unrelated tumors, nucleotide-specific variants are shared in 1.3% (cancer-type agnostic) and in 10% to 13% (cancer-type specific) of cases. The MPS method contextualized shared variants to specify whether patients had a single cancer versus multiple distinct cancers. When multiple tumors were compared from the same patient and an initial clinicopathologic diagnosis was discordant with molecular findings, the MPS anticipated future diagnosis changes in 28% of examined cases.

Conclusions: The use of a novel shared variant framework can provide information to clarify the molecular relationship between compared biospecimens with minimal required input.

Figure 1:. Comparison of distinct tumor exomes from different patients.

a, Nucleotide-specific variant sharing between 10,000 different tumor pairs from TCGA cancer-type agnostic (top) and cancer-type specific (bottom row: 10,000 random comparisons each for 33 different cancer types with replacement). b, Percent of unique tumor pairs per cancer that share a nucleotide-specific mutation. Hotspot (red circle) versus private (black circle) status is defined by Cancer Hotspots.^, Diameter is proportional to the number of mutations with the same shared prevalence. The highest shared nucleotide-specific mutation per cancer type is labeled (amino-acid consequence). All shared variant frequencies are cited in Supplemental Table S1. Sarcoma, kidney renal clear cell carcinoma, kidney chromophobe and mesothelioma are not shown (Table 1).

Figure 2:. Mutational profile similarity (MPS) score of compared tumors from different patients.

a, MPS score derivation. b, Proportion of unique sample pairs (N=204,725) from 32,172 different tumors (one per patient) that share nucleotide-specific mutations detected with MSK IMPACT. The blown-out panel identifies MPS of compared tumors that share one mutation. Each tumor was derived from a different patient, profiled with MSK-IMPACT, and genomic profiles were randomly compared between patients by pairing a tumor only with a cancer of the same type.

Figure 3:. MPS adjusts for background rates of lineage-dependent drivers to estimate shared identity between compared tumors from the same patient.

a-b, Number of shared variants and MPS scores of compared tumors sourced two patients profiled with MSK IMPACT. Months from first diagnosis are shown. c, Comparison of MSK IMPACT genomic profiles of two separate tumors sourced from the same patient (n=2,012 tumor pairs, each from one patient). Datapoint size is proportional to the number of different patients with compared samples that exhibit the same MPS score (y-axis) and number of shared mutations (x-axis). d, Area under the curve (AUC) for the MPS score to detect the same versus different cancers from paired specimens of the same patient. e, Concordance between initial clinicopathologic diagnosis that two separate tumors from a patient represent a single (related) versus distinct (unrelated) cancers and the molecular diagnosis of higher (MPS >= 99%, top) or lower (MPS <99%, bottom) genomic similarity.

Figure 4:. MPS anticipates shared genomic identity in challenging clinical scenarios.

a, Patients clinically diagnosed initially with distinct cancers which share nucleotide-specific variants (n=17). MPS scores and variant number are shown (Supplementary Table S6). Acronyms include: squamous cell carcinomas (SCC), adenocarcinoma (Adeno), hepatocellular carcinoma (HCC), and glioblastoma (GBM). b, MPS application to patient cases where two SCCs may represent secondary primary tumors or a single metastatic disease (Supplementary Table S7). Cases are dichotomized based on the initial clinical diagnosis. c, Shared mutations and MPS of tumors that undergo histologic lineage transformation from an adenocarcinoma to a neuroendocrine tumor.

Figure 5:. Average mutational profile similarity scores of pairs of tumor and plasma sample that share exactly one hotspot versus private nucleotide-specific variant.

Variant hotspot status was classified by Cancer Hotspots^,. Error bars represent 95% confidence intervals of the average Mutational Profile Similarity (MPS). p-values are designated with asterixis and were obtained with Wilcoxon-Mann-U: *p<0.05, **p<0.01, ***p<0.001.