AI-driven analyses of confocal micrographs for biomarker discovery to enable targeted cancer therapy
Dr. Kasturi Mitra (Department of Biology) is leading a groundbreaking research initiative to develop AI-ML-driven prediction models that could revolutionise personalised cancer therapy by identifying novel mitochondria-related biomarkers for targeted treatment strategies.
At the core of this investigation lies the fact that mitochondria, the cellular powerhouses, undergo significant structural and functional alterations in cancer cells that directly impact energy metabolism—a hallmark characteristic of malignancy. The research addresses the critical challenge that not all patients respond uniformly to standard chemotherapeutics, necessitating individualised treatment approaches that account for the distinct genetic and environmental profiles of each cancer patient.
The team has developed a sophisticated computational framework utilising the quantitative mito-SinComp method to analyse confocal micrographs and establish predictive relationships between mitochondrial structure and function. This approach recognises that mitochondrial dynamics, governed by fission and fusion events, along with alterations in the oxidative environment, serve as potential diagnostic and prognostic indicators for cancer progression.
Initial results demonstrate promising predictive capabilities using Random Forest algorithms applied to data from ten experimental groups studying environmental carcinogen impacts. The research has revealed that functional parameters can be accurately predicted from structural features, specifically within network-classified mitochondrial populations. The project’s innovative contribution lies in its potential to transform mitochondrial alterations into quantitative biomarkers that could guide drug screening processes, optimise chemotherapeutic regimens, and enable precise cancer staging. Future developments aim to expand this computational approach to patient-derived cancer cells through hospital collaborations, ultimately creating a robust diagnostic tool for personalized oncology applications.
