Determining mutational burden and signature using RNA-seq from tumor-only samples.

Area of Science:

• Genomics
• Cancer Research
• Bioinformatics

Background:

• Traditional cancer mutational analysis requires paired tumor-normal DNA sequencing, which is not always clinically feasible.
• Investigating tumor RNA sequencing as an alternative can reduce costs and provide gene expression and fusion data.

Purpose of the Study:

• To develop and validate methods for determining mutational burden and signatures from tumor RNA sequencing data.
• To assess the efficacy of RNA-seq in classifying cancer's molecular causes, specifically microsatellite instability (MSI) and DNA polymerase epsilon (POLɛ) defects.

Main Methods:

• Devised supervised and unsupervised learning methods to identify mutational signatures from tumor RNA-seq data.
• Applied methods to TCGA uterine corpus endometrial carcinoma (UCEC) and colorectal cancer (COAD) datasets.
• Developed procedures to filter out germline variants from RNA-seq data.

Main Results:

• RNA-derived mutational burdens significantly associated with MSI and POLɛ status.
• Over 80% of variants explained by COSMIC mutational signatures 5, 6, and 10 (aging, MSI-H, POLɛ).
• Achieved high recall and specificity in classifying MSI and POLɛ status in UCEC, with improved classification in COAD using learned signatures.

Conclusions:

• Presents a novel RNA-seq based method for detecting mutational signatures and burdens.
• Effective germline variant removal enables accurate classification of DNA damage deficiency mechanisms.
• This approach offers a viable, cost-effective alternative for cancer molecular subtyping.