UNISOM: Unified Somatic Calling and Machine Learning-based Classification Enhance the Discovery of CHIP.

Area of Science:

• Genomics
• Computational Biology
• Hematology

Background:

• Clonal hematopoiesis of indeterminate potential (CHIP) is linked to increased risks of hematologic malignancies, cardiovascular disease, and mortality.
• Current CHIP detection methods struggle with low variant allele frequencies (VAFs), often requiring deep targeted sequencing.
• Accurate identification of CHIP mutations is crucial for risk stratification in healthy individuals.

Purpose of the Study:

• To introduce UNISOM, a streamlined workflow for improved CHIP detection from whole-genome and whole-exome sequencing data.
• To enhance the sensitivity of CHIP mutation detection, particularly for low VAFs.
• To provide a scalable solution for CHIP screening in population genomic studies.

Main Methods:

• UNISOM employs a meta-caller for variant detection.
• Machine learning models are integrated to classify variants into CHIP, germline, or artifact categories.
• The workflow is validated on whole-exome and whole-genome sequencing data.

Main Results:

• UNISOM recovered approximately 80% of CHIP mutations identified by deep targeted sequencing in whole-exome data.
• Analysis of whole-genome sequencing data confirmed known CHIP patterns, including gene frequencies, mutation types, and associations with age and smoking.
• UNISOM demonstrated high sensitivity, detecting CHIP mutations with VAFs below 5% in 30% of identified cases.

Conclusions:

• UNISOM offers a sensitive and efficient method for detecting CHIP mutations from standard sequencing data, including low VAFs.
• The workflow facilitates CHIP screening in large-scale population genomic studies.
• UNISOM is freely available, promoting its adoption in research and clinical settings.