Predicting functional bioactivities in fermented milk using deep learning on liquid chromatography-mass spectrometry...

Area of Science:

• Food Science and Technology
• Biotechnology
• Computational Biology

Background:

• Fermented dairy products offer nutritional and health benefits.
• Assessing these functional properties via traditional assays is time-consuming and limits scalability.
• Developing efficient screening methods is crucial for functional food development.

Purpose of the Study:

• To investigate the potential of deep learning (DL) as a rapid and efficient alternative for assessing the bioactivity of fermented milk products.
• To develop and validate a DL model capable of predicting bioactivity scores from metabolomic data.
• To explore the utility of DL in functional food development with limited sample sizes.

Main Methods:

• Liquid chromatography-mass spectrometry (LC-MS) metabolomics was used to analyze 18 fermented milk samples.
• In vitro bioactivity assays (antioxidant, enzyme inhibition, anticancer) were performed.
• A one-dimensional convolutional neural network (1D-CNN) was trained on preprocessed LC-MS data, incorporating data augmentation and regularization.

Main Results:

• The 1D-CNN model achieved a mean absolute error of 0.548 ± 0.089 across all bioactivity predictions, demonstrating strong generalization.
• Principal component analysis revealed distinct clustering of samples based on milk type and fermentation conditions.
• The DL approach successfully predicted functional bioactivity from metabolomic signatures despite a small dataset.

Conclusions:

• Deep learning, combined with robust preprocessing and data augmentation, can accurately predict the functional bioactivity of fermented milk products from metabolomic data.
• This DL-based approach offers a scalable and efficient alternative to traditional bioactivity assays.
• The findings suggest a promising pathway for DL-assisted screening in the development of functional foods.