Abstract C008: Validation of an AI-powered computational chemistry workflow for streamlined drug discovery

Abstract Uncovering druggable oncogenic vulnerabilities and designing novel chemical entities to target them are central priorities in cancer research, especially for advanced malignancies with limited treatment options. However, traditional drug design is time-consuming and resource-intensive. Integrating computational strategies offers a promising solution to accelerate this process. To address this need, we developed DrugAppy, a comprehensive platform that combines artificial intelligence (AI), computational modeling, and medicinal chemistry to enhance the identification of novel cancer therapeutics. DrugAppy uses a hybrid framework that integrates high-throughput virtual screening (HTVS) via molecular docking tools and in-depth evaluation of candidate compounds through molecular dynamics (MD) simulations. Key pharmacological properties—including pharmacokinetics, target selectivity, and predicted bioactivity—are assessed using a mix of public models and proprietary AI algorithms trained on large-scale public datasets. The platform outputs a set of chemical descriptors to support rational compound prioritization. We validated DrugAppy using two oncology-relevant targets: PARP1 and TEAD4. A virtual screening campaign was performed on a chemical library of 4.3 million compounds using a stepwise docking protocol that incorporated standard and high-precision SMINA and GNINA. This approach reduced the initial pool to approximately 1,500 candidates. The top-ranked ligands were then evaluated using MD simulations to determine their binding free energy and stability, which were key descriptors to rank the compounds together with ADMET properties, off-target selectivity, and ML scores. In the PARP1 case study, 9 out of 30 selected compounds were previously reported in the literature to inhibit the target protein. Of the 17 compounds tested in vitro, one molecule demonstrated potent activity (IC50 = 1.31 nM), comparable to the clinically approved PARP1 inhibitor olaparib (IC50 = 1.61 nM). Moreover, the γH2AX assay confirmed target engagement of this compound. In the TEAD4 case study, 1 out of 8 selected compounds exhibited superior in vitro performance (IC50 = 13–18 µM) compared to IK-930 (IC50 = 25–158 µM), a known TEAD family inhibitor. Inhibition of canonical YAP/TAZ-TEAD transcriptional targets suggests that this compound acts as a pan-TEAD inhibitor, with strong binding affinity for both TEAD2 and TEAD4. In both cases, the lead compounds confirmed target engagement and matched or exceeded benchmark efficacy. These findings illustrate the power of the DrugAppy to identify and prioritize novel molecular scaffolds for targeted cancer therapy. The platform enables efficient early-stage drug discovery through the integration of AI, docking, and MD simulations, supporting the development of next-generation cancer therapeutics with improved rationality and speed. Citation Format: Elisa Poyatos-Racionero, Lucía Paniagua-Herranz, Cristian Privat, Carmen Martín-Hernández, Cristina Nieto-Jiménez, Cristina Herrero-Igartua, Ramón García-Escudero, Pawel Gancarski, Corina Lorz, Alberto Ocana. Validation of an AI-powered computational chemistry workflow for streamlined drug discovery [abstract]. In: Proceedings of the AACR-NCI-EORTC International Conference on Molecular Targets and Cancer Therapeutics; 2025 Oct 22-26; Boston, MA. Philadelphia (PA): AACR; Mol Cancer Ther 2025;24(10 Suppl):Abstract nr C008.