GAGA: a new algorithm for genomic inference of geographic ancestry reveals fine level population substructure in Europeans

PLoS Comput Biol. 2014 Feb 20;10(2):e1003480. doi: 10.1371/journal.pcbi.1003480. eCollection 2014 Feb.

Abstract

Attempts to detect genetic population substructure in humans are troubled by the fact that the vast majority of the total amount of observed genetic variation is present within populations rather than between populations. Here we introduce a new algorithm for transforming a genetic distance matrix that reduces the within-population variation considerably. Extensive computer simulations revealed that the transformed matrix captured the genetic population differentiation better than the original one which was based on the T1 statistic. In an empirical genomic data set comprising 2,457 individuals from 23 different European subpopulations, the proportion of individuals that were determined as a genetic neighbour to another individual from the same sampling location increased from 25% with the original matrix to 52% with the transformed matrix. Similarly, the percentage of genetic variation explained between populations by means of Analysis of Molecular Variance (AMOVA) increased from 1.62% to 7.98%. Furthermore, the first two dimensions of a classical multidimensional scaling (MDS) using the transformed matrix explained 15% of the variance, compared to 0.7% obtained with the original matrix. Application of MDS with Mclust, SPA with Mclust, and GemTools algorithms to the same dataset also showed that the transformed matrix gave a better association of the genetic clusters with the sampling locations, and particularly so when it was used in the AMOVA framework with a genetic algorithm. Overall, the new matrix transformation introduced here substantially reduces the within population genetic differentiation, and can be broadly applied to methods such as AMOVA to enhance their sensitivity to reveal population substructure. We herewith provide a publically available (http://www.erasmusmc.nl/fmb/resources/GAGA) model-free method for improved genetic population substructure detection that can be applied to human as well as any other species data in future studies relevant to evolutionary biology, behavioural ecology, medicine, and forensics.

Publication types

  • Research Support, Non-U.S. Gov't

MeSH terms

  • Algorithms*
  • Analysis of Variance
  • Computational Biology
  • Computer Simulation
  • Databases, Genetic / statistics & numerical data
  • Europe
  • Genetic Variation
  • Genetics, Population / statistics & numerical data*
  • Genome, Human
  • Humans
  • Models, Genetic
  • Multigene Family
  • Polymorphism, Single Nucleotide
  • White People / genetics

Grants and funding

This study was funded in part by the Erasmus University Medical Center Rotterdam, the Netherlands Forensic Institute, and the Netherlands Genomics Initiative (NGI)/Netherlands Organization for Scientific Research (NWO) within the framework of the Forensic Genomics Consortium Netherlands (FGCN). AW was supported by Volkswagen Foundation (ref 86042). The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.