Cross-Species and Tissue-Agnostic Prediction of Human Cancer Treatment Response using AI-Powered Cellular
None:
The success of drug development relies heavily on the use of animal models. However, increasing evidence shows that discoveries in these models often fail to translate to human patients. In this study, we developed an AI framework to discovery cross-species and tissue-agnostic cellular morphometric biomarkers (CMBs) as a new avenue to improve translatability. Using this framework, we identified CMBs and CMB tumor subtypes from treatment-naive needle biopsies of mammary tumors in a genetically diverse Erbb2/Neu mouse model, showing significant correlation with responses to docetaxel treatment. These CMBs were then successfully translated to human patients with breast cancer, ovarian cancer or lung cancer, showing superior predictive/prognostic power to biomarkers and/or machine learning systems specifically optimized in human patients. Furthermore, Co-enrichment analysis revealed significantly conserved biological functions associated with CMBs in both mice and humans, such as cell cycle regulation, underscoring their relevance to tumor biology, treatment response and translatability.
