Breed identification using breed-informative SNPs and machine learning based on whole genome sequence data and SNP chip data

Table 3 Identification accuracies (Mean (SE) over 50 replications) when the reference population was genotyped with sequencing and the test population was genotyped with different SNP chips or with sequencing^a

Chip/SEQ	No. MBI SNPs contained^b	Imputation accuracy	Identification accuracy, %
Chip/SEQ	No. MBI SNPs contained^b	Imputation accuracy	KNN	RF	SVM	KSR
50K	20	83.58%	99.69 (0.01)	97.87 (0.03)	98.01 (0.00)	99.16 (0.02)
80K	52	88.52%	99.86 (0.00)	99.19 (0.04)	98.16 (0.00)	99.51 (0.03)
100K	65	89.41%	99.33 (0.01)	98.75 (0.03)	98.01 (0.00)	98.84 (0.02)
150K	91	91.16%	99.65 (0.01)	99.04 (0.03)	98.01 (0.00)	99.19 (0.03)
777K	261	94.36%	99.72 (0.00)	99.15 (0.03)	98.01 (0.00)	99.30 (0.03)
SEQ	2,000	–	99.86 (0.00)	99.24 (0.03)	98.16 (0.00)	99.65 (0.03)

^aThe chip genotypes were imputed to sequence level. The reference population size was 30 individuals per breed and 2,000 most breed-informative SNPs derived by DFI were used
^bNumber of SNPs among the 2,000 most breed-informative (MBI) SNPs derived from the reference population which were contained in the chips
KNN K-Nearest Neighbor, RF Random Forest, SVM Support Vector Machine, KSR, an integration of KNN, SVM and RF

ISSN: 2049-1891