Development of a Novel Non-invasive Metabolomics Assay to Predict Implantation Potential of Human Embryos

S. Cabello-Pinedo,H. Abdulla,S. Mas,A. Fraire,B. Maroto,M. Seth-Smith,M. Escriba,J. Teruel,J. Crespo,S. Munné,J. A. Horcajadas
DOI: https://doi.org/10.1007/s43032-024-01583-y
2024-06-06
Reproductive Sciences
Abstract:Can a set of metabolites present in embryo culture media correlate with embryo implantation? Case-control study in two phases: discovery phase (101 samples) and validation phase (169 samples), collected between 2018 and 2022, with a total of 218 participants. Culture media samples with known implantation outcomes were collected after blastocyst embryo transfer (including both PGT and non-PGT cycles) and were analyzed using chromatography followed by mass spectrometry. The spectra were processed and analyzed using statistical and machine learning techniques to identify biomarkers associated with embryo implantation, and to develop a predictive model. In the discovery phase, 148 embryo implantation biomarkers were identified using high resolution equipment, and 47 of them were characterized. Our results indicate a significant enrichment of tryptophan metabolism, arginine and proline metabolism, and lysine degradation biochemical pathways. After transferring the method to a lower resolution equipment, a model able to assign a Metabolite Pregnancy Index (MPI) to each embryo culture media was developed, taking the concentration of 36 biomarkers as input. Applying this model to 20% of the validation samples (N=34) used as the test set, an accuracy of 85.29% was achieved, with a PPV (Positive Predictive Value) of 88% and a NPV (Negative Predictive Value) of 77.78%. Additionally, informative results were obtained for all the analyzed samples. Metabolite concentration in the media after in vitro culture shows correlation with embryo implantation potential. Furthermore, the mathematical combination of biomarker concentrations using Artificial Intelligence techniques can be used to predict embryo implantation outcome with an accuracy of around 85%.
obstetrics & gynecology,reproductive biology
What problem does this paper attempt to address?