Optimization Study on Non-Invasive Prenatal Testing for Y Chromosome Target Concentration Using Box-Cox Transformation and K-Means Clustering

Abstract

Non-invasive prenatal testing (NIPT) faces challenges in clinical practice, including difficulties in achieving target Y chromosome concentrations in high-BMI pregnant women and the lack of individualized testing timepoints. This study addresses these two critical challenges by constructing a precise nonlinear regression model and developing a timing optimization strategy. To quantify the nonlinear effects of gestational age and BMI on Y chromosome concentration, data preprocessing was performed, including mean imputation for missing BMI values and linear interpolation for gestational age to ensure data stability and continuity. Pearson correlation analysis revealed a strong positive correlation between Y chromosome concentration and gestational age (r=0.72) and a strong negative correlation with BMI (r=−0.58), both identified as critical influencing factors. To capture nonlinear relationships, this study employed Box-Cox transformation combined with stepwise regression. The constructed nonlinear regression model achieved an R² value of 0.875, with both gestational age and BMI exhibiting extremely significant effects (p=0.000), enabling precise prediction of Y chromosome concentration. Building upon this predictive framework, to determine optimal testing timepoints for pregnant women with varying BMIs, K-means clustering (optimal k=5) was employed to categorize female fetus pregnancies into reasonable groups. Clustering results were validated via ANOVA (F=124.7, p<0.001), demonstrating significant differences and clinical interpretability. Subsequently, the optimal testing window with minimal risk was determined using a piecewise risk function, constrained to ensure a pass rate ≥90%. Results indicate that the optimal timing is delayed with increasing BMI, enabling personalized timing recommendations for high-BMI pregnant women.