Abstract
PURPOSE: To evaluate the value of mtDNA copy number measurement in spent culture medium of blastocysts for predicting the chance of implantation following single blastocyst transfer (SBT). METHODS: Copy numbers of mtDNA and genomic DNA (gDNA) were determined using multiplex PCR and NGS and modeled to predict implantation following SBT using a generalized linear model (GLM), generalized additive model (GAM), and extreme gradient boosting (XGBoost). The predictive power of the models was demonstrated and compared with the area under the receiver operating characteristic curve (AUC-ROC). RESULTS: Neither the mtDNA copy number nor the mtDNA/gDNA provided meaningful discriminatory power for prediction in GLM and GAM models. However, higher gDNA quartiles were associated with a negative correlation with pregnancy (OR 0.92, 95% CI 0.85, 1) and an interaction with mtDNA, suggesting that gDNA should not be used to normalize mtDNA copy number. An XGBoost model, which considered both mtDNA and gDNA values, demonstrated an AUC of 0.837 (95% CI 0.800, 0.874). CONCLUSIONS: The mtDNA copy number in spent medium alone may not be a reliable predictor of pregnancy, and dividing mtDNA by gDNA could distort the outcome. Alternatively, a model that makes full use of the interaction of the values may improve the prediction power.