Comparison of the effectiveness of Models 1, 2a and 2b in estimating fetal fraction (FF) based on telomere content.

Model 1, which uses telomere content with (A) or without maternal age (D) , demonstrates low correlation with FF (r = 0.23) and high RMSE (0.043), indicating limited predictive accuracy. The p-values were statistically significant (6.381e-64/1.985e-65). When seqFF was combined with telomere content in Models 2a (C) and 2b (F) , the correlation increased substantially (r = 0.8751 for both). The RMSE decreased to 0.0216 for the seqFF model and 0.0190 for the comboFF model, indicating a strong predictive improvement. Incorporating maternal age did not further enhance correlation but maintained statistical significance (p-value < 1.00e-05 for all comparisons). Models utilizing seqFF alone (Model 7a (B) ) and comboFF alone (Model 7b (E) ) achieved high correlations (r = 0.8723 and r = 0.9056, respectively), demonstrating the baseline predictive power of these models without additional features.