Development of a non-invasive method for selection of human embryos with high implantation potential based on Fourier transform Infrared (FTIR) spectroscopy and artificial neural networks

MARCELO GARCEA; MARCOS HORTON; OSVALDO YANTORNO; ALEJANDRA BOSCH; CECILIA BEATRIZ FÍGOLI; CLAUDIO BISCIOLI

Helsinki

Congreso; Annual Meeting of the European Society of Human Reproduction and Embryology; 2016

ESHRE

Study question:Does FT-IR spectroscopy of spent culture embryo supernatants provide an OMICS?like view of the chemical/biochemical status of the embryo metabolism to be used to evaluate embryo implantation potential?Summary answer:Multivariate analysis of metabolomic spectral patterns of human IVF embryos developed by an artificial neuronal network system could identify an embryo with high implantation potencial.What is known already:The metaboloma is the chemical print produced in a biological system, representing its functional phenotype, and consists of a set of 2500 to 3000 molecules and potential biomarkers. Metabolomic techniques were used as non-invasive approaches for IVF studies analyzing different metabolites in embryo-culture supernatants. Particularly, near infrared spectroscopy (NIR) has been used to evaluate embryo pregnancy ability. Although these results were initially encouraging, a recent randomized clinical trial did not show benefits of NIR over conventional morphological evaluation. FT-IR has been widely used to study the biochemical composition of biological materials. Nevertheless it has never been applied in IVF technology.Study design, size, duration:Prospective cohort study, enrolling unselected patients undergoing IVF or ICSI cycles from November 2012 to December 2014. Four hundred fourteen patients were enrolled, 400 of them reached egg retrieval, 346 had appropriate ova for fertilization an 294 achieved embryo transfer (268 on day 3, 26 on day 5). This is an ongoing collaborative project between a private practice ART clinic and a University Research Center.Participants/materials, setting, methods:Participants with up to 3 previous IVF cycles, and up to 41 years of age were excluded. Embryos were cultured individually in trigas incubators, and 1 or 2 embryos were transferred under conventional clinical and embryological criteria on day 3 or 5. An aliquot of spent culture media was frozen in liquid nitrogen for future spectral acquirement and data analysis. A total of 497 embryos were transferred on day 3, and 38 on day 5. Main results and the role of chance:Five hundred and thirty-five embryos were transferred, and 136 of them implanted (25.4%). After thawing, the sample was centrifuged for oil extraction, droplets were transferred to a ZnSe plate, and vacuum dried for 40 minutes. Due to technical problems with oil extraction in day 5 embryos, only day 3 embryos were finally evaluated. Spectra coming from a total of 1315 embryos, comprising the 497 day 3 transferred embryos and the non-transferred, vitrified embryos, were used to construct the database. Spectra from transferred embryos were classified as Class I (belonging to full implantation embryos, i.e. singletons in sET or twins in DET), Class II (non implanted) and Class III (50%, 1 of 2 embryos implanted). Multivariate methods were applied in order to obtain a model able to discriminate spectra belonging to Class I from Class II supernatants, using Opus 3.1 and 4.5 software (Bruker Optics, Germany), that include the IDENT package for cluster analysis and principal component analysis and NeuroDeveloper®, (Synthon Analytics, Germany) for the development of artificial neural networks applied in the construction of libraries and other databases. This showed a 77% accuracy rate, 80% sensitivity and 76% specificity in predicting implantation.Limitations, reasons for caution:Due to the characteristics of ANNs, needing a huge number of cases to ?learn? during the process, clinical validation of these results still need a larger dataset of cases, to increase accuracy through further analysis of full implantation cases (singletons in sET, or twins after DET)Wider implications of the findings:Addition of other non-invasive embryo selection techniques like morphokinetics could improve the accuracy rate and positive and negative predictive value of metabolomic assessment and help in the construction of predictive models to improve embryo selection and secondarily pregnancy rate.Trial registration number:Non randomized trial.