Multimodal machine learning for modeling infant head circumference, mothers' milk composition, and their shared...

Area of Science:

• Nutritional immunology
• Developmental biology
• Computational biology

Background:

• Understanding the relationship between human milk (HM) and infant development is crucial but complex, often focusing narrowly on individual components.
• Previous research has not fully elucidated the multi-faceted impact of HM on early-life growth and neurodevelopment.

Purpose of the Study:

• To investigate the predictive links between human milk composition and infant head circumference, a proxy for brain development, using advanced machine learning.
• To identify specific HM components and maternal factors associated with infant head growth within a large cohort.

Main Methods:

• Applied multi-modal predictive machine learning to data from 1022 mother-infant dyads in the CHILD Cohort.
• Integrated comprehensive HM data (oligosaccharides, fatty acids, hormones, chemokines) with demographic, health, dietary, and environmental factors.
• Utilized statistical analysis to identify significant associations between HM analytes and head circumference at 3 and 12 months.

Main Results:

• Infant head circumference was significantly predictable using the integrated dataset at both 3 and 12 months of age.
• Human milk's n3-polyunsaturated fatty acid C22:6n3 (docosahexaenoic acid, DHA) was a highly significant predictor of head circumference.
• Maternal fish intake, a primary source of DHA, also showed a significant association with infant head circumference.

Conclusions:

• A systems biology approach successfully identified meaningful relationships between human milk composition and infant brain development.
• The findings validate the predictive statistical models and support novel associations between specific HM components like DHA and infant growth.
• This study provides a foundation for future research into HM's role in neurodevelopment, potentially informing infant nutrition strategies.