Clinically actionable cancer somatic variants (CACSV): a tumor interpreted dataset for analytical workflows

Abstract

Background The recent development and enormous application of parallel sequencing technology in oncology has produced immense amounts of cell-specific genetic information. However, publicly available cell-specific genetic variants are not explained by well-established guidelines. Additionally, cell-specific variants interpretation and classification has remained a challenging task and lacks standardization. The Association for Molecular Pathology (AMP), the American Society of Clinical Oncology (ASCO), and the College of American Pathologists (CAP) published the first consensus guidelines for cell-specific variants cataloging and clinical annotations. Methods AMP–ASCO–CAP recommended sources and information were downloaded and used as follows: relative knowledge in oncology clinical practice guidelines; approved, investigative or preclinical drugs; supporting literature and each gene-tumor site correlation. All information was homogenized into a single knowledgebase. Finally, we incorporated the consensus recommendations into a new computational method. Results A subset of cancer genetic variants was manually curated to benchmark our method and well-known computational algorithms. We applied the new method on freely available tumor-specific databases to produce a clinically actionable cancer somatic variants (CACSV) dataset in an easy-to-integrate format for most clinical analytical workflows. The research also showed the current challenges and limitations of using different classification systems or computational methods. Conclusion CACSV is a step toward cell-specific genetic variants standardized interpretation as it is readily adaptable by most clinical laboratory pipelines for somatic variants clinical annotations. CACSV is freely accessible at (https://github.com/tsobahytm/CACSV/tree/main/dataset).

CACSV: a computational web-sever that provides classification for cancer somatic genetic variants from different tissues

“…In our original work, we covered ten types of tumors including: breast, central nervous system (cns), colon, esophageal, gastric, melanoma, non-small cell lung cancer (nsclc), pancreas, rectal, and small cell lung cancer (sclc) [4]. About three million cancer mutations that are available on COSMIC, intOgen, or the bulk dataset of cBioPortal were collected and classified as previously described.…”

“…We incorporated the clinical recommendations into a new computational method as described previously [4]. The National Comprehensive Cancer Network Clinical Practice Guidelines in Oncology (NCCN Guidelines ® ) were used as the source of the practical clinical evidence.…”

“…Our method was evaluated on a subset of manually reviewed variants and showed a balanced performance on a significantly imbalanced subset [4]. We collected cell-specific mutations from oncology genetic hubs, then filtered germline genetic mutations using the gnomAD database returning around ~ 3 million cancer mutations.…”