Abstract 4383: Inferring transcription activity changes from copy-number and expression data of longitudinally sampled high-grade serous ovarian cancer tumors

Abstract

Abstract To predict patient response and optimal treatment strategies, we present models for integrating copy-number and expression data sampled longitudinally in high-grade serous ovarian cancer. Experimental procedures: We analyzed 91 prospectively collected samples from 60 high-serous ovarian cancer patients, collected before and after chemotherapy. Whole-genome and/or whole-transcriptome sequencing was performed on each sample, in order to quantify copy-number profiles and the transcriptomic activity. Additionally, 235 ovarian cancer samples containing both whole-genome and whole-transcriptome data from primary tumors of the The Cancer Genome Atlas (TCGA) cohort were used for validation. Results: Chemotherapy exerts both microenvironmental and phenotypic changes on the patient tumors, which manifest in systematic variation in the sequenced profiles. Copy-numbers are often adjusted for a normal population, whereas transcriptomes are not. We report here a novel computational method to adjust the expression data and show that this improves the association between expression profiles and patient survival. Using the adjusted expression data, we constructed co-expression modules for 473 transcription factors (TFs), whose DNA recognition motif is enriched in the upstream sequences of their coexpressed genes, and used a network inference procedure to predict the latent TF activity by modeling the relationships between copy-number and expression data, and the latent TF activity. Compared to expression data, the advantage is that the model integrates information from the (regulatory-wise) neighboring copy-number alterations and expression data. With the above analysis, we identified various TFs whose predicted activity significantly differs between patient groups, such as good/poor treatment response and between the treatment naive and relapsed samples. Several TFs, such as PPAR-alpha, EPAS-1, ELF-1, and GATA-3 were validated to have a similar, significant association between their predicted activity and patient survival in the TCGA cohort. Conclusions: Our methodology allows quantitative estimation of pathway and TF activity in individual patient samples, which facilitates analysis and comparison of patient tumor evolution during the treatment. The identified TFs might confer sensitivity/resistance chemotherapy, and, as shown, allow better predicting patient survival. Consequently, our results allow ranking putative intervention strategies for overcoming ovarian cancer chemoresistance in future validation experiments. Citation Format: Antti Häkkinen, Kaiyang Zhang, Johanna Hynninen, Sakari Hietanen, Kaisa Huhtinen, Olli Carpén, Rainer Lehtonen, Sampsa Hautaniemi. Inferring transcription activity changes from copy-number and expression data of longitudinally sampled high-grade serous ovarian cancer tumors [abstract]. In: Proceedings of the Annual Meeting of the American Association for Cancer Research 2020; 2020 Apr 27-28 and Jun 22-24. Philadelphia (PA): AACR; Cancer Res 2020;80(16 Suppl):Abstract nr 4383.