Universität zu Köln

Streck, Adam ORCID: 0000-0002-7302-0147 and Schwarz, Roland F ORCID: 0000-0001-9155-4268 (2025). CNSistent integration and feature extraction from somatic copy number profiles. GigaScience, 14. pp. 1-11. Oxford University Press. ISSN 2047-217X

PDF giaf104.pdf Bereitstellung unter der CC-Lizenz: Creative Commons Attribution. Download (2MB)

Abstract

[Artikel-Nr.: giaf104] Background: Most cancers exhibit somatic copy number alterations (SCNAs)—gains and losses of variable regions of DNA. SCNAs play a key role in cancer adaptation through modulation of gene expression, deletion of tumor suppressor genes, or amplification of oncogenes. Systematic analysis of SCNAs is now a routine task in both the clinic and research and can help identify novel cancer genes, improve our understanding of cancer gene regulation, and enable us to accurately reconstruct cancer phylogenies. However, to conduct such analyses, SCNA profiles have to be integrated between samples, patients, and cohorts—often a nontrivial task, for which dedicated toolkits are lacking. Results: To fill this gap, we developed CNSistent, a Python package for imputation, filtering, consistent segmentation, feature ex- traction, and visualization of cancer copy number profiles from heterogeneous datasets. We demonstrate the utility of CNSistent by applying it to the following publicly available cohorts: The Cancer Genome Atlas, Pan-Cancer Analysis of Whole Genomes, and TRAcking Cancer Evolution through therapy (Rx). We compare the effect of sample preprocessing and different segmentation and aggregation strategies on cancer type and subtype classification tasks using various classification models. We also evaluate how well a classifier trained on one cohort generalizes to another. Lastly, we introduce 2 segment-based peak and outlier scores to investigate relationships between segments, between samples, and between cancer types. Using these scores, we investigate non–small cell lung cancer samples, highlighting that SOX2 amplification is the dominant copy number alteration in lung squamous cell carcinoma and the main distinction to lung adenocarcinoma. Conclusions: CNSistent is a general-purpose toolkit for integrated processing of SCNA profiles across many patients and cohorts. It is available at https://bitbucket.org/schwarzlab/cnsistent. The Research Resource Identifier for CNSistent is SCR_027025.

Item Type:	Article
Creators:	Creators Email ORCID ORCID Put Code Streck, Adam UNSPECIFIED https://orcid.org/0000-0002-7302-0147 UNSPECIFIED Schwarz, Roland F UNSPECIFIED https://orcid.org/0000-0001-9155-4268 UNSPECIFIED
URN:	urn:nbn:de:hbz:38-811419
Identification Number:	10.1093/gigascience/giaf104
Journal or Publication Title:	GigaScience
Volume:	14
Page Range:	pp. 1-11
Number of Pages:	11
Date:	13 September 2025
Publisher:	Oxford University Press
ISSN:	2047-217X
Language:	English
Faculty:	Faculty of Medicine
Divisions:	Faculty of Medicine > Weitere > Centrum für integrierte Onkologie (CIO)
Subjects:	Medical sciences Medicine
Uncontrolled Keywords:	Keywords Language cancer ; data processing ; SCNA, deep learning ; cancer classification English
['eprint_fieldname_oa_funders' not defined]:	Publikationsfonds UzK
Refereed:	Yes
URI:	http://kups.ub.uni-koeln.de/id/eprint/81141