Back

Explainable deep learning on 7500 whole genomes elucidates cancer-specific patterns of chromosomal instability

al-Badri, M. A.; Cross, W. C.; Barnes, C. P.

2024-03-13 cancer biology
10.1101/2024.03.08.584160 bioRxiv
Show abstract

Chromosomal instability leads to an increased rate of chromosome or part-chromosome copy gains or losses. Referred to as chromosome copy aberrations (CNAs), these mutations are highly prevalent in cancer cells and contribute to abnormal genome structure, genetic diversity within a tumour, drug resistance, and evolution to metastatic disease. Within each cancer type, it remains unclear how CNAs define each particular malignant phenotype, yet some key events have been characterized. Here, we present a novel deep learning method, DeepCNA, and apply it as part of an exploration of CNAs present across 7,500 whole genomes and 13 cancer types. DeepCNA is an explainable AI approach, that reveals both established and novel loci that contribute to cancer type. It can be used as a diagnostic tool in situations such as cancer of unknown primary site, and it can discover CNAs that are prognostic. It also has several applications for researchers, including drug target and biomarker discovery.

Matching journals

The top 7 journals account for 50% of the predicted probability mass.

50% of probability mass above

"Similar papers" are the closest papers from that journal in the model's embedding space. They show what the match is built on, but the ranking comes mostly from a classifier over the whole training set, not from these examples alone.