CLEMENT: genomic decomposition and reconstruction of non-tumor subclones

TL;DR

CLEMENT is a new algorithm that accurately identifies and reconstructs subclones in non-tumor genome sequencing data.

Contribution

CLEMENT introduces a novel algorithm for non-tumor subclone decomposition using EM optimization strategies tailored for normal samples.

Findings

CLEMENT outperformed cancer decomposition algorithms in estimating clone numbers and variant-clone membership.

Testing on normal tissue sequencing confirmed accurate subclone identification from different cell types.

Clone-level analysis revealed new insights into mutational burden and signatures in normal tissues.

Abstract

Genome-level clonal decomposition of a single specimen has been widely studied; however, it is mostly limited to cancer research. In this study, we developed a new algorithm CLEMENT, which conducts accurate decomposition and reconstruction of multiple subclones in genome sequencing of non-tumor (normal) samples. CLEMENT employs the Expectation-Maximization (EM) algorithm with optimization strategies specific to non-tumor subclones, including false variant call identification, non-disparate clone fuzzy clustering, and clonal allele fraction confinement. In the simulation and in vitro cell line mixture data, CLEMENT outperformed current cancer decomposition algorithms in estimating the number of clones (root-mean-square-error = 0.58–0.78 versus 1.43–3.34) and in the variant-clone membership agreement (∼85.5% versus 70.1–76.7%). Additional testing on human multi-clonal normal tissue sequencing confirmed the accurate identification of subclones that originated from different cell types. Clone-level analysis, including mutational burden and signatures, provided a new understanding of normal-tissue composition. We expect that CLEMENT will serve as a crucial tool in the currently emerging field of non-tumor genome analysis.

Graphical Abstract

Figures