Abstract 4891: Discovery of novel biomolecular condensate drug targets in oncology using in silico predictive tools

Abstract

Abstract Biomolecular condensates play a crucial role in maintaining cellular homeostasis by regulating various cellular processes, such as signaling, gene expression, DNA repair, and stress response. Dysregulation of condensate function has been implicated in the onset and progression of several cancer types, highlighting the potential of targeting condensates as a therapeutic approach. To establish a pipeline for discovering and prioritizing new oncology targets that function via biomolecular condensates, we developed in silico models that link protein condensate formation to its sequence. Leveraging publicly available data and neural network based protein language models, we built predictive models for protein phase separation and for their localization into heteromolecular condensates. With these models, we mapped the condensation character of the full human proteome, including disease variants of genes/proteins. Our results suggested that approximately half of the proteome has the potential to localize into condensate systems. We identified a broad array of chromosomal rearrangements that lead to pathological condensation (e.g., EML4-ALK fusions in lung cancer) and several instances where disease specific isoforms suppressed condensation (e.g., BARD1 mutations in breast cancer and DAXX mutations in melanoma). To identify cases where commonly occurring missense mutations lead to abnormal condensation we analyzed omics data from patients and healthy volunteers, focusing on two mechanisms: (i) condensation due to protein overexpression and (ii) altered condensation characteristics of disease variants compared to wild type protein. We investigated 18 TCGA oncology indications and identified several targets where mutations were predicted to cause aberrant condensation. The identified targets presented a wide range of biological classes, with transcription factors, co-activators, and epigenetic modulators being the most prominent. We prioritized CTNNB1 (βcatenin) for experimental validation and successfully demonstrated both the phase separation ability of the reconstituted protein and the presence of CTNNB1 condensates in colon carcinoma cells with mutated WNT pathway. Notably, many of our predicted targets contain extended intrinsically disordered regions (IDRs), which have remained challenging to target using conventional drug discovery approaches due to their lack of a well-defined 3D structure. To identify compounds that modulate IDR containing proteins we are utilizing our PhaseScan™ platform, which can screen tens of thousands of compounds against our predicted targets within condensates and measure their modulation potential. Subsequent cellular profiling of active compounds in the predicted disease models will allow us to validate the predictions made by our in silico models and ultimately unlock drugging novel targets within biomolecular condensate. Citation Format: Kadi L. Saar, Marius Rebmann, Mohammed Kanchwala, Seema Qamar, Prathima Radhakrishnan, Jasmine Cornish, Julia Doh, Jerome Cattin, Assaf Rotem, Mahmoud Ghandi, Andrew Seeber, Richard C. Centore, Sarah A. Teichmann, Tuomas Knowles, Shilpi Arora. Discovery of novel biomolecular condensate drug targets in oncology using in silico predictive tools [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 4891.