10.11588/data/TT4VM3Peil, BarbaraBarbaraPeilInstitute of Medical Biometry and Informatics, University of Heidelberg, Heidelberg, GermanyKabisch, MariaMariaKabischMolecular Genetics of Breast Cancer, German Cancer Research Center (DKFZ), Heidelberg, GermanyFischer, ChristineChristineFischerInstitute of Human Genetics, University Hospital Heidelberg, Heidelberg, GermanyHamann, UteUteHamannMolecular Genetics of Breast Cancer, German Cancer Research Center (DKFZ), Heidelberg, GermanyLorenzo Bermejo, JustoJustoLorenzo Bermejo0000-0002-6568-5333Institute of Medical Biometry and Informatics, University of Heidelberg, Heidelberg, GermanyTailored Selection of Study Individuals to be Sequenced in Order to Improve the Accuracy of Genotype Imputation [Source Code]heiDATA2018Medicine, Health and Life Sciencesgenotype imputationsequencingreference panelHapMapstatistical depthLorenzo Bermejo, JustoJustoLorenzo BermejoInstitute of Medical Biometry and Informatics, University of Heidelberg, Heidelberg, Germany2018-07-202018-10-1510.1002/gepi.218734700345431720application/pdftype/x-r-syntaxtype/x-r-syntax1.0The addition of sequence data from own‐study individuals to genotypes from external data repositories, for example, the HapMap, has been shown to improve the accuracy of imputed genotypes. Early approaches for reference panel selection favored individuals who best reflect recombination patterns in the study population. By contrast, a maximization of genetic diversity in the reference panel has been recently proposed. We investigate here a novel strategy to select individuals for sequencing that relies on the characterization of the ancestral kernel of the study population. The simulated study scenarios consisted of several combinations of subpopulations from HapMap. HapMap individuals who did not belong to the study population constituted an external reference panel which was complemented with the sequences of study individuals selected according to different strategies. In addition to a random choice, individuals with the largest statistical depth according to the first genetic principal components were selected. In all simulated scenarios the integration of sequences from own‐study individuals increased imputation accuracy. The selection of individuals based on the statistical depth resulted in the highest imputation accuracy for European and Asian study scenarios, whereas random selection performed best for an African‐study scenario. Present findings indicate that there is no universal ‘best strategy’ to select individuals for sequencing. We propose to use the methodology described in the manuscript to assess the advantage of focusing on the ancestral kernel under own study characteristics (study size, genetic diversity, availability and properties of external reference panels, frequency of imputed variants…).