Machine Learning-enhanced Genomic Profiling for Precision Paediatric Blood Transfusion Compatibility

Abstract Background: Paediatric patients who require frequent blood transfusions are at increased risk of developing immune complications such as alloimmunisation and haemolytic reactions. These risks arise in part because traditional serological methods often fail to detect critical antigen mismatches. This study addresses this challenge by integrating genomic profiling with machine learning (ML) to enhance the prediction of transfusion compatibility. To design and evaluate an ML-powered genomic profiling system that accurately predicts transfusion compatibility in paediatric patients. Methodology: A hybrid model combining Bayesian inference and convolutional neural networks was developed to analyse next-generation sequencing data across 50 erythrocyte antigen systems. A graph neural network was used to integrate genomic, clinical and human leucocyte antigen data to assess transfusion risk. Compatibility thresholds were dynamically adjusted based on individual patient profiles. The system was trained on a dataset of 50,000 paediatric cases and operates in real time with an inference latency of <500 ms using TensorRT optimisation. Results: Compared to conventional and rule-based genomic methods, the proposed system demonstrated improved performance: 0.97 sensitivity in variant detection, 0.93 area under the curve in compatibility prediction and high precision in predicting haemolytic (0.88) and alloimmunisation risks (0.91). It also identified 142 novel variants specific to paediatric patients that are undetectable through serology. Notably, neonates and oncology patients experienced the greatest benefit in compatibility accuracy and risk stratification. Conclusion: This study introduces a next-generation tool for safer, more precise blood transfusions in paediatric care. By combining in-depth genetic analysis with powerful ML, the system marks a significant step forward in tailoring transfusion practices to the unique needs of paediatric patients.