Convolutional neural networks to reveal resistant phenotypes behind the complex genotypes of ovarian cancer

High-grade serous ovarian cancer (HGSOC), which is the most aggressive type of ovarian cancer, is characterized by the mutation of gene TP53 and extensive copy number variations (CNVs).

HGSOC tumors typically show an initial favourable response to standard treatments, however they often acquire resistance and relapse. Currently, there is a need for new therapeutic approaches to combat emerging drug-resistant subpopulations.

Although, the scarcity of common targetable oncogenic mutations has complicated the development of directed therapies, the study of CNVs offers a promising opportunity to find new mechanisms of resistance and develop alternative treatments, as it has been described how CNVs can model clonal fitness and therapeutic resistance in other types of cancer.Here, I will explore the impact of the CNVs on the treatment response of HGSOC patients and their distal effects on the transcriptome using convolutional neural networks.

This complex machine learning model will be trained with the largest available longitudinal cohort of HGSOC samples at the single-cell resolution and will reveal which CNVs, and their specific combinations, have a relevant role in shaping the HGSOC tumours upon treatment.

Employing a systems biology approach I will identify convergent phenotypes within these relevant CNV profiles and then validate their effects on treatment using data from both external cohorts and drug-treated patient derived organoid models.

This novel approach enables revealing resistance mechanisms driven by complex genotypes, and thus allows finding specific vulnerabilities to combat emerging resistance in HGSOC.