Abstract / Summary
Traditional Non-Invasive Prenatal Testing (NIPT) for Trisomy 21 (Down Syndrome) is typically limited to binary diagnostic outcomes, neglecting the complex, systemic nature of the syndrome and its long-term physiological impacts. In this paper, a novel, end-to-end Clinical Decision Support System (CDSS) is proposed, which integrates machine learning and transcriptomics to provide a holistic framework spanning from initial prenatal diagnosis to lifelong predictive monitoring. Initially, the platform employs a Random Forest algorithm, reinforced by SHapley Additive exPlanations (SHAP) for Explainable AI (XAI), to robustly predict Trisomy 21 risk using maternal blood cfDNA parameters (e.g., Chr21 read ratio, fetal fraction, GC bias). For high-risk cases, the system triggers a molecular impact mapping module utilizing Rbased RNA-Seq differential gene expression (DGE) analysis to monitor the overexpression of critical chromosome 21 genes, such as APP and DYRK1A. This molecular data feeds into a multi-output predictive model to project the longitudinal risk of secondary pathologies, including early-onset Alzheimer's disease and leukemia. Furthermore, the platform introduces a pioneering non-invasive monitoring module that tracks daily behavioral trends via a relational database and analyzes acoustic biomarkers using librosa to detect vocal hesitation and cognitive load—key early indicators of dementia. By unifying cfDNA diagnostics, multi-omics risk projection, and acoustic biomarker analysis into a single microservice-oriented dashboard, this framework transitions Trisomy 21 management from a static diagnostic event to a continuous, explainable, and cost-effective predictive continuum.