PlaNET

PlaNET infers sample ethnicity from placental DNA methylation profiles to control for population stratification in epigenome-wide association studies.

Key Features:

• Ethnicity classification: Predicts major classes of self-reported ethnicity/race including African, Asian, and Caucasian with reported accuracy of 0.938 and a kappa statistic of 0.823.
• Platform compatibility: Uses data from the Infinium Human Methylation 450k BeadChip (HM450k) and is compatible with the EPIC platform.
• Genetic ancestry correlation: Produces ethnicity probabilities that are highly correlated with genetic ancestry from genome-wide SNP arrays (over 2.5 million SNPs) and with 50 ancestry informative markers.
• Genetically informative CpG sites: Classification relies on 1,860 HM450k microarray sites, of which 955 are linked to nearby genetic polymorphisms, integrating methylation and genetic signals.
• Placenta optimization: Specifically optimized for placental samples and reported to outperform existing methods in assessing population stratification across diverse ethnic groups.

Scientific Applications:

• Control of population stratification in placental DNAme studies: Provides an ethnicity covariate to reduce confounding in placental epigenome-wide association studies.
• Ethnicity inference when self-report is missing or unreliable: Infers ethnicity as a discrete classification or as continuous probabilities for downstream analyses.
• Ancestry estimation without genotyping: Offers an alternative means to assess genetic ancestry when genotyping markers are unavailable.

Methodology:

PlaNET employs an Elastic Net classifier trained on methylation data from 509 placental samples across five cohorts using 1,860 HM450k sites, 955 of which are linked to nearby genetic polymorphisms.