GAprediction

GAprediction predicts gestational age at birth from cord blood DNA methylation profiles measured on the Illumina HumanMethylation450 BeadChip.

Key Features:

• Data Source: Uses Illumina HumanMethylation450 BeadChip CpG data from cord blood for genome-wide CpG coverage.
• Training and Validation: Developed using the Norwegian Mother and Child Birth Cohort (MoBa) with MoBa 1 (n=1068) for training and MoBa 2 (n=685) for validation (total n=1753).
• Statistical Modeling: Builds a statistical model to predict gestational age with adjustments for covariates including estimated cell type proportions and applies Bonferroni correction for multiple testing.
• Comparative Analysis: Compares model predictions to ultrasound- and LMP-derived gestational ages and reports superior agreement with ultrasound-based estimates.
• Distinctive Biomarkers: Identifies specific CpG sites and associated genes linked to gestational age that are distinct from those used in chronological age prediction models.

Scientific Applications:

• Enhanced Accuracy: Produces gestational age predictions that improve on LMP-based estimates and align more closely with ultrasound-based estimates.
• Research Insights: Reveals DNA methylation patterns and associated genes related to fetal development and gestational age.
• Comparative Studies: Enables analysis of methylation-profile differences between ultrasound- and LMP-based gestational age estimates.

Methodology:

Uses Illumina HumanMethylation450 cord blood data; trains a statistical model on MoBa 1 to identify CpGs associated with gestational age adjusted for covariates; validates the model on MoBa 2 and compares predictions to ultrasound and LMP estimates; identifies CpGs and associated genes distinct from chronological age models.