Abstract 893: Studying the impact of CNVs on expression at single-cell resolution in HGSOC using autoencoders

Abstract High-grade serous ovarian cancer (HGSOC) is one of the most lethal gynecological malignancies. Lack of common targetable oncogenic mutations has complicated the development of directed therapies to combat emerging resistance. This malignancy is mainly characterized by the mutation of gene TP53, which promotes genome instability for the emergence of extensive copy number variations (CNVs). However its impact on gene expression at the single-cell level is not well understood. In this study, we aim to investigate the effect of CNVs on transcriptomic signatures by taking advantage of variational autoencoders (VAE) ability for dimensionality reduction, unsupervised learning and feature extraction. The use of VAEs is becoming more popular for the analysis of scRNAseq data, and scVI is one of the most versatile VAE applications performing wide variety of tasks. Here, we used single-cell RNA sequencing (scRNA-seq) data from 90 longitudinal samples of 64 HGSOC patients and inferred CNVs in each cell using inferCNV, an established computational pipeline. Then, we modified scVI algorithm to allow the VAE to reconstruct CNVs from a latent space originated from gene expression profiles and viceversa. Our models were capable of reconstructing CNV profiles accurately from expression data and also remove batch effect. From these models we could observe how, after the integration of genomic information, the latent clusters produced from the transcriptomic space were influenced by the amplifications or deletions of certain genomic regions. Moreover, some of these clusters were characterized by the alteration of important oncogenes in HGSOC, such as KRAS and CCNE1, and allowed us to focus on the transcriptomic consequences of their amplifications. With the results from the approach presented here, we gained a more comprehensive picture of the impact of genomic alterations on HGSOC. As future work, we plan to validate of this results on external cohorts and link the identified signatures to clinically relevant features such as prognosis or chemotherapy response. Citation Format: Matias Marin Falco, Teemu Närhi, Erdogan Pekcan Erkan, Johanna Hynninen, Anna Vähärautio. Studying the impact of CNVs on expression at single-cell resolution in HGSOC using autoencoders [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2024; Part 1 (Regular Abstracts); 2024 Apr 5-10; San Diego, CA. Philadelphia (PA): AACR; Cancer Res 2024;84(6_Suppl):Abstract nr 893.