planet

planet provides R functions to predict biological variables from placental DNA methylation data generated from Infinium arrays.

Key Features:

• Ethnicity prediction: Predicts ethnicity/ancestry as a discrete or continuous variable using 1860 HM450K microarray sites, with accuracy of 0.938 and kappa of 0.823 for major classes of self-reported ethnicity/race (African, Asian, Caucasian).
• Gestational age prediction: Infers gestational age from placental DNA methylation data.
• Cell composition estimation: Estimates cell composition directly from placental DNAm data based on reference methylome data from four placental cell populations (trophoblasts, Hofbauer cells, endothelial cells, stromal cells).
• Platform compatibility: Compatible with Infinium Human Methylation 450k BeadChip array (HM450k) and EPIC platform data.

Scientific Applications:

• Population stratification assessment: Addresses confounding by population stratification in placental DNAme association studies when self-reported ethnicity information is missing and genotyping markers are unavailable.
• Placental methylome characterization: Enables cell-specific characterization of the placental methylome by estimating cell composition in mixed whole placental tissue samples.
• Cross-gestation analysis: Supports analysis of cell-specific DNA methylation dynamics across gestation from first trimester to term pregnancies.

Methodology:

The ethnicity classifier uses an elastic net approach (PlaNET: Placental DNAme Elastic Net Ethnicity Tool) trained on 509 placental samples from five cohorts. Cell composition estimation uses reference methylome data from sorted placental cell populations profiled on the Illumina EPIC methylation array.