476 Aryastha bio Solutions, a phase zero platform that integrates patient-derived tumor organoids and neural networks to predict immune checkpoint responses for colorectal cancers

Background Cancer immunotherapy has evolved over decades into a frontline treatment for multiple malignancies. Colorectal cancer (CRC), an early-onset and high-mortality cancer type, demonstrates varied patient responses to these therapies. This variability underscores an unmet clinical need for adaptive strategies that dissect resistance mechanisms, guide patient selection, inform combinatorial treatment approaches, and optimize clinical trial design.Methods Leveraging multimodal in silico analysis—spanning DNA methylation, genomics, and transcriptomics data from The Cancer Genome Atlas (TCGA)—we developed machine-learning classifiers to predict CRC responses to FDA-approved immune checkpoint inhibitors (ICIs). These classifiers informed the design of Phase Zero ‘mini-trials’ using a patient-derived tumor organoid (PDTO) biobank. PDTOs, established from surgically resected CRC tissues, were expanded up to five passages. They underwent genomic and histopathological validation and were assessed for proliferation (Ki67), membrane integrity, viability (WST-8), and apoptosis (Caspase-3c).Results On a population scale, we constructed CRC-specific gene regulatory networks categorized by tumor-infiltrating lymphocytes (TILs), tertiary lymphoid structure (TLS), and tumor mutational burden. This modeling captured the immunogenomic diversity of CRC and illuminated both convergent and divergent resistance patterns to ICIs. These predictions guided the Phase Zero PDTO mini-trial. PDTOs were co-cultured with either autologous or allogeneic immune cells (PBMCs, TILs) and treated with ICIs—alone or combined with drugs identified via in silico modeling as potentially synergistic. Using high-throughput live imaging, we tracked organoid diameter, structural integrity, and metabolic activity. Responder organoids displayed dynamic morphological and metabolic changes, whereas resistant ones remained largely unchanged. Immune infiltration and tumor-immune interaction dynamics were also mapped. By integrating genetic and phenotypic data, our model improved prediction accuracy of ICI response and resistance. These organoid-based outcomes were then compared with patient response data, to validate our platform’s predictive fidelity.Conclusions Our integrated response modeling demonstrates the value of translatable preclinical platforms in bridging bench-to-bedside gaps. Aryastha Bio Solutions offers a clinically relevant Phase Zero drug-screening platform that enables both drug developers and clinicians to:• Recapitulate phenotypic heterogeneity of CRC• Probe tumor-immune interactions• Identify responder signatures• Evaluate therapeutic candidates• Inform rational combinatorial strategies• Tailor immunotherapy regimens at the individual-patient levelBy faithfully mirroring in vivo tumor heterogeneity and immune context, our platform supports smarter therapy selection, resistance mechanism elucidation, and optimized trial design—advancing precision immunotherapy.