Machine learning-based genome-wide association analysis to construct a clinical decision model for severe neonatal jaundice.

Background: Early identification of severe unconjugated hyperbilirubinemia is critical to prevent bilirubin encephalopathy and long-term neurological damage. The etiology of neonatal jaundice is complex. This study aims to identify genetic variants associated with severe neonatal jaundice (SNJ) through a genome-wide association approach and to evaluate their potential for risk prediction. Methods: We conducted a genome-wide association study using whole-exome sequencing data from 155 SNJ cases and 160 controls without SNJ. Genetic and clinical variables were integrated using a LASSO-based machine learning approach to assess their importance in classifying SNJ. Causal inference methods were applied to evaluate the relationship between identified single-nucleotide polymorphisms (SNPs) and SNJ. Results: In our cohort, SNJ was associated with increased erythrocyte count and hemoglobin (Hb) concentration. A positive correlation between erythrocyte count and Hb was also observed. Seventeen SNPs were found to be significantly associated with total blood erythrocyte count. A missense mutation in the gene haptoglobin-related protein ( Conclusions: We applied machine learning causal inference to GWAS data and identified potential erythroid-related genetic factors, including the