NAToRA, a relatedness-pruning method to minimize the loss of dataset size in genetic and omics analyses

Abstract

Genetic and omics analyses frequently require independent observations, which is not guaranteed in real datasets. When relatedness can not be accounted for, solutions involve removing related individuals (or observations) and, consequently, a reduction of available data. We developed a network-based relatedness-pruning method that minimizes dataset reduction while removing unwanted relationships in a dataset. It uses node degree centrality metric to identify highly connected nodes (or individuals) and implements heuristics that approximate the minimal reduction of a dataset to allow its application to large datasets. NAToRA outperformed two popular methodologies (implemented in software PLINK and KING) by showing the best combination of effective relatedness-pruning, removing all relatives while keeping the largest possible number of individuals in all datasets tested and also, with similar or lesser reduction in genetic diversity. NAToRA is freely available, both as a standalone tool that can be easily incorporated as part of a pipeline, and as a graphical web tool that allows visualization of the relatedness networks. NAToRA also accepts a variety of relationship metrics as input, which facilitates its use. We also present a genealogies simulator software used for different tests performed in the manuscript.

Unappreciated Subcontinental Admixture in Europeans and European Americans: Implications for Genetic Epidemiology Studies

“…Based on these IBD probabilities, we calculated the pairwise kinship coefficient ( Φ ij ) as a function of IBD-sharing, Φ ij = 1/2 δ 2 ij + 1/4 δ 1 ij . We modeled the genetic relationships among individuals as networks 52 , in which pairs of individuals were linked if they had a Φij threshold ≥ 0.0884 ( i.e ., first- and second-degree relatives 53 ). Then, we excluded related individuals using the maximum clique graph approach to minimize sample loss 52 .…”

“…We modeled the genetic relationships among individuals as networks 52 , in which pairs of individuals were linked if they had a Φij threshold ≥ 0.0884 ( i.e ., first- and second-degree relatives 53 ). Then, we excluded related individuals using the maximum clique graph approach to minimize sample loss 52 . We performed unsupervised principal components analysis 13 and unsupervised ADMIXTURE analysis 17 on the European reference data.…”

“…We performed association analysis using linear mixed models implemented in GENESIS 58 . Our analyses were focused on unrelated European Americans, with relatedness determined by the maximum clique graph approach 52 . Models were adjusted for the genetic relationship matrix as a random effect to account for variance components and the four first principal components (significantly associated with the outcome) and/or locus-specific ancestry as fixed effects.…”

X-Chromosome Association Study in Latin American Cohorts Identifies New Loci in Parkinson Disease

“…First, we removed individuals with missing covariates and variants that are: (i) located in structural variants using TriTyper [33], (ii) duplicated, (iii) monomorphic, (iv) potential probe sites (defined as variants in which the probe may have variable affinity due to the presence of other SNPs within 20 bp and with MAF above 1%), or (v) have failed the Hardy-Weinberg Equilibrium (HWE) exact test (p < 10 -5 in controls). Next, we removed individuals with > 10% and variants with more than 5% of missing genetic data and inferred the relatedness between included individuals using KING [34]; those at greater than a second degree level (kinship coefficient > 0.0884) were removed using NAToRA [35] (Figure S2). Unlike the Le Guen et al pipeline, samples were not removed based on ancestry.…”