Multivariate Independent Component Analysis Identifies Patients in Newborn Screening Equally to Adjusted Reference...

Area of Science:

• Biochemistry
• Computational Biology
• Genetics

Background:

• Newborn screening (NBS) relies on reference ranges from healthy populations to detect inborn errors of metabolism (IEMs).
• Current methods use quantile statistics of biomarker concentrations and ratios, which may not capture complex metabolic interdependencies.

Purpose of the Study:

• To evaluate the utility of multivariate independent component analysis (ICA) for analyzing NBS data.
• To explore the structural characteristics of ICA-derived scores in the context of IEM detection.

Main Methods:

• Independent component analysis (ICA) was applied to a large dataset (10,213 controls, 77 patients) from a routine NBS program (2013-2022).
• A discovery/validation phase approach was used, with the same ICA model applied to both phases.
• Results were analyzed by evaluating ICA scores against standard deviation (sd) levels, identifying outlying patient samples.

Main Results:

• Patient samples from 7 distinct IEMs were clearly identified as outliers (5sd) from control groups in both discovery and validation phases.
• Two IEMs with single patients were separated at the 3sd level in the discovery phase.
• A significant association between neonatal birth weight and one latent variable was observed.

Conclusions:

• Multivariate ICA is a promising tool for analyzing NBS data and detecting IEMs.
• ICA has the potential to be integrated into the decision-making process of NBS programs.
• Interpretation of ICA score structures can aid in identifying metabolic abnormalities.