An assessment of Bayesian and multinomial logistic regression classification systems to analyse admixed individuals

• Published in: Forensic science international. Genetics supplement series Vol. 4; no. 1; pp. e63 - e64
• Main Authors: McNevin, D., Santos, C., Gómez-Tato, A., Álvarez-Dios, J., de Cal, M. Casares, Daniel, R., Phillips, C., Lareu, M.V.
• Format: Journal Article
• Language: English
• Published: Elsevier Ireland Ltd 2013
• Subjects: Biogeographical ancestry; Multinomial logistic regression; Phenotype prediction; Structure; Multinomial logistic regression; Phenotype prediction; Structure; Biogeographical ancestry
• ISSN: 1875-1768; 1875-175X
• DOI: 10.1016/j.fsigss.2013.10.032

Multinomial logistic regression (MLR) has been applied to the prediction of hair and eye colour. Here we apply it to the prediction of biogeographical ancestry (BGA) in a test set of 1092 admixed and non-admixed genotypes from the 1000 Genomes Project using a training set of 571 non-admixed genotypes from the HGDP CEPH cell line panel. Predicted BGAs are consistent with those of Structure, a naïve Bayesian classifier.