Benchmarking reference-based cellular deconvolution algorithms to predict cell proportions

Ayesha A. Malik

doi:10.1016/j.compbiolchem.2026.109370

Benchmarking reference-based cellular deconvolution algorithms to predict cell proportions

2026-09-15

0

PRE

AI

DOI：10.1016/j.compbiolchem.2026.109370

Original

Original request for help

Save

Abstract

En 中文

• Benchmarked 10 reference-based deconvolution methods across diverse conditions. • Statistical and Bayesian methods consistently outperformed deep learning methods. • Top methods were robust to sequencing depth, with little gain beyond 10⁵ reads. • Signature reconstruction mirrored proportion estimates, supporting downstream use. • Cross-dataset breast cancer validation recovered meaningful biological patterns.

Keywords:

Cellular deconvolution

Cellular fraction prediction

Cell-type-specific gene expression profile

scRNA-seq

Benchmarking reference-based cellular deconvolution algorithms to predict cell proportions

Abstract

Journal

Organization

Cited Papers