dryclean

dryclean applies robust principal component analysis (rPCA) to de-noise GC- and mappability-corrected read depth data and improve detection of somatic copy number alterations (SCNAs) by removing "waviness" caused by technical and biological confounders.

Key Features:

• Robust foreground detection: Uses rPCA to separate true biological SCNA signals from noisy coverage fluctuations in read depth data.
• Background subtraction: Removes background noise and waviness from read depth to isolate SCNA-relevant signal.
• Platform compatibility: Applicable to whole genome sequencing (WGS) and targeted sequencing platforms.

Scientific Applications:

• Improved SCNA detection: Mitigates replication timing–driven waviness in WGS read depth to enhance detection of biologically relevant SCNAs relative to existing algorithms.
• Enhanced sensitivity in relapse detection: In silico tumor dilution experiments demonstrate up to tenfold improvement in sensitivity for relapse detection.

Methodology:

Employs a robust PCA-based algorithm to de-noise genomic coverage, uses a panel of normal samples (PON) to characterize biological and technical noise in read depth, and operates on GC- and mappability-corrected read depth data (obtainable via fragCounter).