Non-invasive prediction of embryo ploidy status using metabolomic profiling and machine learning

Can a non-invasive approach integrating metabolomic profiling of spent embryo culture medium (SECM), matrix-assisted laser desorption/ionization tandem time-of-flight mass spectrometry (MALDI-TOF MS) and machine learning identify distinct metabolic signatures predictive of embryo ploidy status?

A total of 120 SECM samples were analysed using MALDI-TOF MS and subsequently used to train various machine learning models. These models were then tested to identify the one with the highest accuracy. The samples had previously undergone preimplantation genetic testing for aneuploidies (PGT-A) through next-generation sequencing. In this retrospective study, participants provided consent in a consecutive series, the only eligibility requirement being that PGT-A had to be performed prior to the MALDI-TOF MS analysis.

The results demonstrate high accuracy, sensitivity and reproducibility across samples, with minimal batch effects observed between clinics. In addition to detecting potential biomarkers associated with euploidy, the machine learning model demonstrated a comparable performance compared with current assessment methods. The LightGBM model used produced a 91% accuracy with an average sensitivity of 92.16% and an average specificity of 97.30%.

The approach discussed here is cost-effective, automatable and scalable method for high-throughput screening. While further validation with larger cohorts is ongoing, this method shows significant promise as a practical, non-invasive alternative to traditional embryo assessment tools. Integration with current clinical practices could enhance embryo selection accuracy, ultimately aiming to improve implantation and live birth outcomes in IVF.