Confounder-adjusted analysis of statin selectivity across cancer cell lines identifies a colorectal-enriched SLC45A4 signal

Read full text Summarise with VIRAT
Authors: Takayuki Sakuma
Cited by: 0
References: 28

Summary

No summary yet. VIRAT can read this paper and write a plain-language summary of its aim, methods, key findings and limitations.

Summarise with VIRAT Free to try with your Science Catalogs account.

Abstract

Abstract Large-scale pharmacogenomic screens enable systematic searches for genetic determinants of drug sensitivity, but cross–cell-line analyses are confounded by lineage structure and drugfree proliferation rates. We integrate DepMap genome-scale CRISPR dependency profiles with PRISM viability responses and estimate covariate-adjusted associations between gene dependency and a statin selectivity phenotype using cross-fitted double machine learning (DML). The analysis recovers mevalonate-pathway regulators (MVK, UBIAD1) and identifies SLC45A4 as a colorectal-enriched candidate association. To assess robustness, we validate findings in independent AUC datasets, conduct leave-one-drug-out sensitivity analyses, and benchmark against empirical-null drug-bundle and gene-level negative controls. In colorectal lines, the SLC45A4 signal aligns with sterol-biosynthesis-related co-dependency and colorectal-specific shifts in polyunsaturated cholesteryl esters and putrescine in CCLE metabolomics. As positive controls, expected EGFR- and BRAF-matched drug responses were reproduced in independent datasets. These results reduce concern that the SLC45A4 association is explained solely by cross-dataset integration bias and support further experimental testing.

Fields of Study & Subjects

Cancer Lipids and Metabolism Computational Drug Discovery Methods Cancer Genomics and Diagnostics

Publication details

Journal / source
Research Square
Publication date
26 March 2026
Language
EN
Document type
Preprint