整合组学驱动的数字化身与患者来源的实验模型以加速精准肿瘤学

Precision oncology increasingly depends on linking patient-specific molecular profiles to experimentally validated therapeutic response models. ValiRx and Inaphaea BioLabs are developing advanced patient-derived functional cancer models, while TwinEdge Bioscience is building large-scale Digital Avatar collections that mechanistically represent individual tumours across genomics, transcriptomics, proteomics, regulatory-network dynamics, and drug-response modules. By combining these complementary capabilities, we established a closed-loop translational framework for predicting, testing, and refining precision-treatment strategies, integrating (i) omics profiling of ValiRx's patient-derived cancer models, and (ii) construction of mechanistic Digital Avatars using TwinEdge's modelling engine. Methods: Avatars are formed by integrating gene expression patterns, pathway activity states, inferred regulatory network, and compound-response modules. We generated digital avatars by integrating matched multi-omics datasets - transcriptomic, genomic, proteomic, and phenotypic layers - into large-scale, mechanistic network models that capture cell-state dynamics. Each individual avatar was then embedded within a population of thousands, allowing systematic comparison to identify subgroups that respond to a given intervention through shared mechanistic signatures. These mechanistically aligned avatars were then analysed to uncover repurposing opportunities and novel biomarker candidates. Results: Across breast, ovarian, and colorectal cancer models, the Digital Avatars were shown to faithfully recapitulated tumour-specific regulatory features, including pathway activation, metabolic rewiring, and stress-response signatures. Avatar-based predictions revealed compound-specific vulnerabilities across multiple patient-derived models and uncovered previously uncharacterised mechanisms underlying differential responses to both targeted and broad spectrum agents. For certain models, observations pointed to the loss of regulation around VEGF-A, network rewiring around DNA/damage and some loss of c-Myc regulatory feedback. Together, these findings highlight the likelihood of a partial or full response, and allow suggestion of potential therapeutic intervention that would convert partial responders into full responders. Conclusion: The TwinEdge-ValiRx translational program demonstrates the power of integrating computational Digital Avatars with patient-derived functional models. This combined framework enhances mechanistic interpretability, improves drug response prediction, and accelerates preclinical decision making. These early results form a concrete proof-of-concept that Avatar based loop dynamics capture treatment specific regulatory rewiring and can guide downstream therapeutic evaluation.