Surface-enhanced Raman spectroscopy-based metabolic fingerprinting of embryo culture medium for improved IVF embryo...

Accurate assessment of embryo developmental competence is essential for improving pregnancy outcomes in in vitro fertilization (IVF). To reduce the subjectivity of conventional morphological evaluation and explore complementary metabolic information, we investigated an adjunctive, non-invasive strategy using surface-enhanced Raman spectroscopy (SERS) to analyze spent embryo culture medium (SCM). SERS spectra from 94 clinical embryos (52 high-quality, 42 low-quality) were analyzed using principal component analysis-linear discriminant analysis (PCA-LDA), support vector machine (SVM), partial least squares-discriminant analysis (PLS-DA), and least absolute shrinkage and selection operator-PLS-DA (Lasso-PLS-DA) models. The Lasso-PLS-DA model showed the highest discriminatory performance among the tested models, with an area under the receiver operating characteristic curve (AUC) of 0.766, compared with PCA-LDA (AUC = 0.723), SVM (AUC = 0.716), and PLS-DA (AUC = 0.688). Comparative spectral analysis of transferred embryos preliminarily suggested associations between clinical pregnancy outcomes and spectral features related to phenylalanine, proteins, lipid-related bands, and energy metabolism. These findings indicate that SERS combined with machine learning may provide auxiliary metabolic information for embryo selection. The moderate performance and limited sample size, including few pregnancy cases, indicate the need for further validation in larger independent cohorts with robust endpoints such as preimplantation genetic testing for aneuploidy (PGT-A) and live birth outcomes.