A Multi-Omic Big Data and Artificial Intelligence Framework for Predicting Patient-Specific Efficacy and Lineage Commitment in Stem Cell Therapies for Autism Spectrum Disorder

• Abilly Elly

  Texas University

  Author

Autism Spectrum Disorder (ASD) presents a significant clinical challenge due to its heterogeneous presentation and complex genetic architecture, with traditional therapeutic approaches often failing to account for individual patient variability in treatment response. Despite advances in stem cell therapy and genomic medicine, no validated predictive framework exists to forecast patient-specific outcomes or direct lineage commitment in stem cell-based interventions for ASD. This study addresses this critical gap by developing and validating a multi-omic big data and artificial intelligence framework for predicting therapeutic efficacy in personalized stem cell therapies. A comprehensive dataset of 704 individuals was analyzed, integrating genomic, transcriptomic, epigenomic, and clinical data through an ensemble machine learning approach combining LightGBM, Random Forest, Neural Networks, and XGBoost with a Stacking Ensemble meta-learner. The framework achieved superior predictive performance with an ROC-AUC of 0.9989 and F1-score of 0.9125 in ASD classification, while Neural Networks demonstrated exceptional recall (95.76%) suitable for early screening applications. Key predictive biomarkers were identified, enabling the stratification of patients into distinct treatment-response clusters. The proposed framework provides a replicable, data-driven approach for personalized treatment planning in ASD, with significant implications for clinical decision support, healthcare resource allocation, and the advancement of precision medicine in neurodevelopmental disorders. These findings pave the way for the clinical translation of AI-guided stem cell therapies, though prospective validation in real-world settings remains essential.

• PDF

Copyright (c) 2026 Abilly Elly (Author)

This work is licensed under a Creative Commons Attribution 4.0 International License.