TSomVar: a tumor-only somatic and germline variant identification method with random forest.

Area of Science:

• Genomics
• Cancer Research
• Bioinformatics

Background:

• Somatic variants are crucial in cancer development and require accurate identification for research.
• Identifying somatic variants is challenging, especially without paired normal samples due to sample-specific variations.

Purpose of the Study:

• To develop a robust tumor-only method for identifying somatic and germline variants in cancer samples.
• To overcome the limitations of existing methods that require paired normal samples.

Main Methods:

• Developed TSomVar, a tumor-only variant identification tool utilizing the random forest algorithm.
• Integrated genotype imputation, read-mapping, and functional annotation for sample-specific variant datasets.
• Trained TSomVar on genomic variant data from colorectal cancer, hepatocellular carcinoma, and melanoma.

Main Results:

• TSomVar demonstrated superior performance in somatic variant detection compared to existing tumor-only tools.
• Achieved higher accuracy and recall rates for somatic variants in colorectal cancer and melanoma datasets.
• Successfully identified germline variants in tumor samples with high accuracy.

Conclusions:

• TSomVar offers a reliable solution for somatic variant identification in tumor-only samples.
• The method facilitates and potentially revolutionizes somatic variant exploration in cancer genomics.
• TSomVar enhances the capability to study cancer genomes, particularly when normal samples are unavailable.