PAN-CANCER INTEGRATIVE HISTOLOGY-GENOMIC ANALYSIS VIA INTERPRETABLE MULTIMODAL DEEP LEARNING
Richard J. Chen, Ming Y. Lu, Drew F. K. Williamson, Tiffany Chen, Jana Lipková, Muhammad Shaban, Maha Shady, Mane Williams, Bumjin Joo, Zahra Noor, Faisal Mahmood · Journal of Pathology Informatics · 2022
linearities reflect tradeoffs where one species benefits more than the other during good periods and suffers more (or does less well) than the other during less good periods, be the periods stochastic or seasonal.Methods: For this project, we 1) establish that the conditions in tumors are conducive for this mechanism (angiogensisis, hypoxia), 2) empirically explore biomarkers indicative of the two phenotypes (HIF-1; GLUT-1; CA IX; CA XII), 3) and compare these strategies in ecology in nature to the cancer histologies to provide cross-disciplinary insights into this mechanism of coexistence.Results: Biomarker quantification revealed coexisting cancer cells with high and low expression occurring side by side at 150um spatial scale.CAIX, CAXII, GLUT1 and HIF1 quantification yielded their respective quadrats had 27 versus 20 (57%), 22 versus 38 (33%), 42 versus 73 (37%), and 11 versus 91 (89%) cancer cells showing high versus low expression.This pattern of coexistence was manifest across 25 IDC BrCa patients.Conclusion: Temporal and spatial variation is the cause and consequence of intratumoral heterogeneity.We have quantified and confirmed how these variations drive cancer subpopluations to coexist and thrive which may provide meaningful insights into disease recurrence and treatment resistance, but of which profoundly impact prognoses.