CONICET | Buscador de Institutos y Recursos Humanos

ObjectiveThe increase of pregnancy rates in in-vitro fertilization is a continuous challenge that involves both research field and clinical application. New non-invasive technologies for embryos selection are emerging. The field of metabolomics uses the spent culture media (SCM) as a source of information that enables to infer the status of the embryo. The purpose of this work was to predict blastocyst potential to generate biochemical pregnancy using metabolomic profiling of SCM based on Fourier Transform Infrared Spectroscopy (FTIR).DesignRetrospective.Materials and MethodsTwenty-seven embryos elegibles to be transferred to 19 patients were individually cultured until blastocyst stage in G1plus and G2plus (Vitrolife, Goteberg, Sweden) under 6.5% CO2 and 6% O2 (Ksystem). From each culture drop, 40 μL of spent media and blank were collected and their chemical spectra, range between 400 and 4000 cm-1 (2 cm-1 resolution) were obtained through a FTIR-microscope (iN10 Nicolet, Thermo Scientific, USA). Biochemical pregnancy was confirmed two weeks after embryo transfer.Embryo?s spectral data were analyzed under R language. The first and second derivatives were estimated and data was assessed in the five following intervals: I) 900-1200 cm-1 (proteins and carbohydrates), II) 1300-1500 cm-1 (carbohydrates, proteins and lipids), III) 1500-1800 cm-1 (amide I, amide II and proteins), IV) 2800-3000 cm-1 (lipids), and V) 3000-4000 cm-1 (water and carbohydrates). In order to differentiate embryos based on the biochemical pregnancy outcome, a linear classifier was built using partial-least-squares discriminant analysis. For assessing the performance of the developed model, typical classifier?s metrics (accuracy, sensitivity, specificity and F1-score) were calculated under a leave-one-out cross-validation strategy.ResultsWhen analyzing the partial-least-squares components of the SCM spectrums from blastocysts that generated 15 negative and 12 positive biochemical pregnancies, a differential spectral pattern was found and two separable clusters were obtained. Data belonging to the intervals I and IV were found informative for building the classifier. In the validation stage, the model reached 74% accuracy (79% sensitivity, 69% specificity) with 76% F1-score.ConclusionsOur experiment shows that biochemical pregnancy can be predicted by metabolomic profiling of SCM at blastocyst stage. Embryos that result in biochemical pregnancy show different spectral profiles when compared with embryos that do not achieve it. The metabolomic differences in carbohydrate, proteins and lipid regions were meaningful to discriminate blastocyst with potential to implant. The method could be used in the future to enhance non-invasive embryo selection for improve success in ART.

enviar mensaje