Identification of Concentration Drivers for Male Fetal Free DNA Based on Multiple Linear Regression and Correlation Analysis

Abstract

Addressing the challenge that Y-chromosome concentration in male fetuses during non-invasive prenatal testing is influenced by multiple maternal physiological and technical factors, this study utilized data from 1,082 clinical samples to construct a data-driven model for identifying and quantifying concentration drivers. The study first converted text-based gestational age uniformly into days using a gestational age conversion function. Categorical variables such as mode of delivery and parity were encoded using one-hot encoding. Height and weight indicators with severe multicollinearity were excluded using variance inflation factor to ensure independence of model inputs. During modeling, ordinary least squares established a multiple linear regression model. Spearman's rank correlation coefficient and hierarchical clustering were employed to multidimensionally analyze the association characteristics between 22 independent variables and Y chromosome concentration. Results showed the adjusted R² of this regression model was 0.468, with an F-statistic of 40.61, passing the significance test. The study identified the proportion of filtered reads, the proportion of duplicate reads, and X chromosome concentration as core influencing factors. It also quantified the positive promoting effect of gestational age and the negative diluting effect of body mass index. This research provides scientific quantitative evidence for clinically evaluating test reliability and optimizing sampling timing.