Inference of mutational status, loss of heterozygosity, and clonality in tumor-only data
Year: 2018
Type: other
Abstract: Recent advances in clinical sequencing technologies have resulted in unprecedented access to the genomes of individual tumors. High-depth, hybrid-capture-based assays aim to identify somatic mutations in cancer cells for accurate diagnosis and treatment. However, most clinical-grade implementations lack patient-matched germline DNA and additional analysis is needed to infer variants’ mutational status. For instance, when a potentially pathogenic mutation in a tumor suppressor gene is detected, it is imperative to determine whether it is germline or somatic and resolve any evidence for the loss of the wild-type copy, which may be pertinent for treatment efficacy. Genomic heterogeneity in both tumor and non-tumor cell populations also confounds distinguishing sub-clonal tumor alterations from those possibly originating from the non-tumor component in the tumor microenvironment. Therefore, as the use of tumor-only sequencing assays increases, systematic interpretation of clinical-grade data is necessary to accurately describe the genomics of a single tumor. Here, we present LOHGIC (LOH-Germline Inference Calculator), developed on a model-selection scheme using Akaike Information Criterion weighting. LOHGIC finds the most consistent model for mutations’ germline-versus-somatic status, and infers loss of heterozygosity (LOH), mutated allele’s copy-number, and cancer cell fraction, incorporating inherent biases in clinical sequencing and sample purity estimation. We used LOHGIC to assess BRCA1/2 alterations in 1,636 solid tumors and evaluated its results with genomic testing data, demonstrating 93% accuracy, 100% precision, and 96% recall. This analysis highlighted a differential tumor spectrum associated with BRCA1/2 mutations under LOH: germline BRCA1 mutations were exclusively found in women and were breast cancer or of mullerian origin, whereas germline BRCA2 mutations led to a wider spectrum of cancers in both sexes. Moreover, inferring clonality raised the hypothesis that certain sub-clonal mutations did not reflect alterations arising in cancer cells, but might be present in hematopoietic cells infiltrating the tumor microenvironment. Indeed, sequencing peripheral blood and macrodissected lymphocytes showed that 79% of sub-clonal TET2 and DNMT3A mutations found in the original specimens were not tumor mutations, but instead were detected due to the presence of clonal hematopoiesis of indeterminate potential (CHIP). Analyzing 113,079 solid tumors from 21 cancer types further indicated that many sub-clonal alterations in CHIP-associated genes detected in solid tumor specimens possibly arose from their presence in admixed hematopoietic elements. Our results demonstrate that inference of mutational signatures and dissection of heterogeneity can generate diagnostic hypotheses that when clinically tested, lead to improved prognostication, ensuring that treatment strategies are correctly focused on tumor alterations.
Institution: Rutgers, The State University of New Jersey
Language: English
Subfield: Cancer Research
Domain: Life Sciences
Open Access status: gold
OpenAlex ID: W2891820803