Tangent normalization for somatic copy-number inference in cancer genome analysis

Area of Science:

• Genomics
• Cancer Research
• Bioinformatics

Background:

• Somatic copy-number alterations (SCNAs) are crucial in cancer development.
• Systematic noise in sequencing and array data complicates SCNA inference for cancer genomics.
• The Cancer Genome Atlas (TCGA) initiative generated extensive tumor and normal sample data.

Purpose of the Study:

• To introduce and describe the Tangent normalization method for SCNA inference.
• To present Pseudo-Tangent, a modification for scenarios with limited normal samples.
• To improve the accuracy of SCNA detection in cancer genome analyses.

Main Methods:

• Developed the Tangent normalization method using a linear combination of normal samples as a reference for each tumor.
• Implemented Pseudo-Tangent for denoising by comparing tumor profiles when normal samples are scarce.
• Applied these methods to over 10,000 tumor-normal sample pairs from TCGA.

Main Results:

• Tangent normalization significantly increases signal-to-noise ratios (SNRs) compared to conventional methods for both SNP array and whole-exome sequencing (WES) data.
• Tangent and Pseudo-Tangent reduce noise with minimal impact on the underlying signal, outperforming other analysis steps.
• These methods enable more accurate inference of SCNAs from DNA sequencing and array data.

Conclusions:

• Tangent and Pseudo-Tangent normalization methods enhance the accuracy of SCNA detection in cancer genomics.
• These broadly applicable methods improve data quality for both SNP array and WES analyses.
• Tangent is integrated into GATK4's copy-number pipeline and is publicly available.