FinGeneAI: An Explainable Artificial Intelligence Framework for Predicting Genomic Healthcare Costs Using Multi-Omics Data

Abstract

The multi-omics analysis and utilization of artificial intelligence have become common tools in biomedical research, aiding in the understanding of complex diseases and precision healthcare. To date, most of the clinical studies on genomics have centered on cancer diagnosis, prognosis, prediction, or discovery of biomarkers, but the link between genomic patterns and the burden of health care costs has not been well studied. In this research, we present an explainable machine learning framework, FinGeneAI, for the prediction of categories of cost-burden in healthcare for genomic data related to diseases. The framework involves data preprocessing, feature selection, supervised modeling, model performance analysis, and finally, interpretability of influential biomedical features using Explainable AI. The transcriptomic models are developed and evaluated using computational data sources which are publicly available, such as TCGA/UCI data derived from RNA-seq. Results show that molecular features offer informative predictive signals to classify the level of healthcare burden, and explainability methods facilitate transparent interpretation of important molecular features contributing to the burden. The study also finds the potential to go beyond disease prediction, to estimate a healthcare burden assessment using genomic and transcriptomic information. This work introduces a computational connection between molecular data analysis, predictive modeling and cost-burden explainable decision support for future precision medicine planning..