Combined use of principal component analysis and random forests identify population‐informative single nucleotide polymorphisms: application in cattle breeds

Abstract

The genetic identification of the population of origin of individuals, including animals, has several practical applications in forensics, evolution, conservation genetics, breeding and authentication of animal products. Commercial high-density single nucleotide polymorphism (SNP) genotyping tools that have been recently developed in many species provide information from a large number of polymorphic sites that can be used to identify population-/breed-informative markers. In this study, starting from Illumina BovineSNP50 v1 BeadChip array genotyping data available from 3711 cattle of four breeds (2091 Italian Holstein, 738 Italian Brown, 475 Italian Simmental and 407 Marchigiana), principal component analysis (PCA) and random forests (RFs) were combined to identify informative SNP panels useful for cattle breed identification. From a PCA preselected list of 580 SNPs, RFs were computed using ranking methods (Mean Decrease in the Gini Index and Mean Accuracy Decrease) to identify the most informative 48 and 96 SNPs for breed assignment. The out-of-bag (OOB) error rate for both ranking methods and SNP densities ranged from 0.0 to 0.1% in the reference population. Application of this approach in a test population (10% of individuals pre-extracted from the whole data set) achieved 100% of correct assignment with both classifiers. Linkage disequilibrium between selected SNPs was relevant (r(2) > 0.6) only in few pairs of markers indicating that most of the selected SNPs captured different fractions of variance. Several informative SNPs were in genes/QTL regions that affect or are associated with phenotypes or production traits that might differentiate the investigated breeds. The combination of PCA and RF to perform SNP selection and breed assignment can be easily implemented and is able to identify subsets of informative SNPs useful for population assignment starting from a large number of markers derived by high-throughput genotyping platforms.

Penalized classification for optimal statistical selection of markers from high-throughput genotyping: application in sheep breeds

“…Sottile, Sardina, Mastrangelo, Di Gerlando, Tolone, Chiodi and Portolano Penalized multinomial regression and stability selection v. principal component analysis and random forest Penalized multinomial regression and stability selection procedure is a new strategy used for assigning animals to a breed. In order to compare our approach with other previously reported strategies and to test its efficiency in assigning individuals, PCA and RF strategy (Bertolini et al, 2015) were also used with the real data. With respect to the two first ranking SNP panels (MDGI and MAD for 48 and 96 SNPs), the OOB errors in the test population were 4.09% and 2.03%, respectively, while the misclassification error rates for the validation population were both 2.86%.…”

“…In order to compare our approach to those previously reported, another mixed strategy was considered (Bertolini et al, 2015). In particular, PCA and RF was used to discover a new SNP panel able to discriminate among the breeds For each autosome, the top 20 SNPs were selected and merged together, leading to a final panel of 520 markers.…”

“…Four different SNP panels were created selecting the first 48 and 96 SNPs from the MDGI and the first 48 and 96 from the MAD, respectively. This SNP panels' size was chosen considering the practical possibilities to develop multiplex SNP panels containing a reduced number of SNPs for field applications (Bertolini et al, 2015). For each of the four reduced panels, a new RF was fitted and the corresponding out-of-bag (OOB) error rate was calculated.…”

“…The identification of individuals' breed/population of origin offers has several practical applications in livestock and is useful in different biological contexts, such as management of livestock genetic resources for breed confirmation, estimation of hybridization level and authentication of brand products that are produced using only a few particular breeds or populations (Wilkinson et al, 2011;Bertolini et al, 2015). Moreover, assignment of individuals to a specific breed is very important both for biodiversity purposes and products traceability, especially when the phenotypic differentiation among breeds is difficult (Tolone et al, 2012).…”

“…Many clustering algorithms have been developed employing population genetic data to assign individuals to clusters (Jakobsson and Rosenberg, 2007). Several statistical methods were used to determine which genetic markers contain the most information to discriminate among populations (Rosenberg, 2005;Wilkinson et al, 2011), such as the combined approach of principal component analysis (PCA) and random forest (RF) (Bertolini et al, 2015), multivariate canonical discriminant analysis (Dimauro et al, 2013), the statistic delta (Shriver et al, 1997), and Wright's F st (Bowcock et al, 1994). While all these methodologies yielded reduced marker panels useful for breed identification, the power of assignment varied among analysis methods.…”