Non-Invasive Embryo Quality Assessment via Matrix-Optimized Untargeted LC-MS Metabolomics of Spent Embryo Culture Media and Weighted Ensemble Machine Learning

Authors

Gan, H., Wang, X., Tang, F., Ibrahim, H., Chen, X., Xie, P., Zhang, S., Lin, G., Zeng, J., Chu, H., Zhang, S.

Abstract

Spent embryo culture media (SECM) presents significant challenges for untargeted LC-MS metabolomics due to limited sample volume, high salt content, and abundant proteins. Despite growing interest in SECM metabolomics for non-invasive embryo quality assessment, systematic optimization of sample preparation for this unique matrix remains largely unexplored. Here we performed a comprehensive evaluation of six extraction solvent compositions across five solvent-to-sample volume ratios, followed by systematic screening of eleven reconstitution solvent compositions, using human serum as a reference biofluid. Our results establish that optimal sample preparation is fundamentally matrix-dependent: SECM required 7xvolume of 50% acetonitrile for extraction and 40% acetonitrile for reconstitution to maximize metabolome coverage, whereas serum required 10xvolume of 100% methanol and 100% water. Comparative metabolomics revealed that SECM contained a higher proportion of non-polar species, whereas serum showed a higher proportion of polar compounds, directly rationalizing the observed matrix-specific requirements. Applying the optimized workflow to 120 clinical samples, we identified 102 differential metabolites distinguishing euploid from aneuploid embryos, with significant enrichment in fatty acid metabolism, mitochondrial {beta}oxidation, and sphingolipid metabolism, alongside contributions from amino acid and TCA cycle metabolites. To translate these signatures into a predictive tool, we developed a weighted ensemble machine learning model that integrates seven complementary classifiers through an accuracy-weighted voting mechanism. This ensemble approach achieved excellent performance in discriminating aneuploid from euploid embryos (AUC = 0.977, 100.0% specificity, 89.6% sensitivity), substantially outperforming any single classifier. Among 72 euploid embryos stratified by morphological quality (good, fair, poor), the model further revealed progressive metabolic deterioration from mitochondrial energy deficiency to broader lipid dysregulation as quality declined, with AUCs of 0.944, 0.889, and 0.943, respectively. This systematic optimization provides a robust sample preparation protocol for SECM metabolomics, while the weighted ensemble framework establishes a generalizable approach for leveraging metabolomics data in clinical decision support.

Preprint server: bioRxiv
The authors list and abstract were imported from bioRxiv on 05 Aug 2026.

Stats

1-terrible, 9-excellent. How would you rate this preprint? Sign in in to submit your rating.

Post a comment

You need to be signed in to post comments. You can sign in here.

Comments

There are no comments yet.