Phenotype forecasting with SNPs data through gene-based Bayesian networks

BMC Bioinformatics

Table 1 Classification performance on the test sets.

Model	SNP based		Meta-variable based		Haplotype based		Majority Classifier
Classification Accuracies (%) and K statistics	CA	K-stat	CA	K-stat	CA	K-stat	CA
Sampling test 1	55.71	0.09	64.28	0.26	57.14	0.12	51.43
Sampling test 2	55	0.07	59.28	0.16	53.57	0.04	51.43
Sampling test 3	63.57	0.25	67.86	0.34	55	0.07	51.43
Sampling test 4	62.14	0.22	65.72	0.29	49.29	-0.04	51.43
Sampling test 5	58.57	0.15	64.28	0.26	57.85	0.13	51.43
Mean values on test sets	58.99	0.16	64.28	0.26	54.57	0.06	51.43
95% Confidence Interval	54.28–63.72		60.36–68.2		50.34–58.80
Standard Deviation	3.8		3.16		3.4
Standard Error	1.7		1.41		1.52

The table summarizes the results obtained by repeating 5 times a random sampling hold-out scheme in which 75% of the dataset (216 affected and 203 unaffected individuals) was employed as training set and the remaining 25% as test set (72 affected and 68 unaffected individuals). In particular, the table shows the classification accuracies obtained on the test sets by the single-SNP BN, the meta-variable BN and the haplotype BN, the accuracies of the majority classifier and the k-statistics.

ISSN: 1471-2105