Machine learning uncovers previously undetectable genetic deletions in cancer panel sequencing.

PTEN

Extreme Gradient Boosting model detecting focal PTEN deletions and their zygosity, missed by the bin-count CNV approach of TSO500

Within the MASTERsg precision oncology program, the authors developed a machine learning pipeline for deletion detection and zygosity evaluation on panel data, focusing on the tumor suppressor gene PTEN, whose loss confers sensitivity to mTOR/Akt inhibition. Standard pipelines such as TSO500 use bin count averaging for copy number variation detection and often overlook focal events. An Extreme Gradient Boosting model trained on PTEN-deleted and wild-type tumor samples identified focal deletions missed by TSO500 and CNVkit, associated with concordant PTEN protein loss by immunohistochemistry. The model detected PTEN deletions in 33 of 204 samples classified as wild-type by TSO500; 75% (15/20) of stainable cases showed absent or reduced PTEN protein expression.

Synthesis written by Geno'X. For the full original abstract, please refer to the source publication.

This is somatic work on tumor tissue: its scope is precision oncology and the search for targetable alterations, not constitutional diagnosis. The methodological message goes beyond PTEN: focal deletions escape the bin-based CNV approaches of panels, a blind spot that also exists for germline panels. The abstract reports neither treatment response in the 33 reclassified patients nor orthogonal validation other than immunohistochemistry, assessable in only 20 cases.

Analysis by Dr Thibaut Benquey