Table 1.

False positive and False negative rates for the binary classification, regression and MLVO in test set. The MLVO has systematically lower FP and FN values. When using the same data as the MLVO in combination with a cutoff to classify, the results are worse than Using an SVM.

	BIMAS		SVM		MLVO		SVM +cutoff
	Sens.	Spec.	Sens.	Spec.	Sens.	Spec.	Sens.	Spec.
A*0101	0.7692	0.8385	0.8462	0.9142	1	0.9736
A*0201	0.8499	0.872	0.7949	0.8298	0.935	0.9116
A*0301	0.6765	0.8744	0.8824	0.8367	1	0.9292	0.792	0.844
A*1101	0.3929	0.9439	0.878	0.8788	0.9767	0.9304	0.781	0.828
A*2402	0.8837	0.7615	0.907	0.8796	0.9821	0.916	0.866	0.932
A*3101	0.5556	0.9197	0.8889	0.8912	1	0.9348	0.782	0.83
A*6801	1	0.8166	0.963	0.8995	1	0.9376	0.903	0.95
B*0702	0.8095	0.9129	0.925	0.8853	0.9048	0.9536
B*0801	0.7297	0.9298	0.8108	0.8454	0.8919	0.954
B*1501	0.9531	0.817	0.8906	0.9176	1	0.9464	0.79	0.84
B*2702	1	0.9109	1	0.939	1	0.9712
B*2705	1	0.6215	0.9922	0.9566	1	0.9634
B*3501	0.902	0.7996	0.902	0.8812	0.9804	0.9568
B*3701	0.92	0.7156	0.84	0.8601	0.96	0.9562
B*3901	1	0.818	0.9726	0.9423	1	0.9672
B*40	1	0.8016	1	0.9304	1	0.963
B*4001	1	0.9028	1	0.9375	1	0.974
B*4403	1	0.8548	0.9048	0.9072	1	0.9536	0.928	0.847
B*5101	0.9697	0.8312	0.9394	0.9154	1	0.9602	0.853	0.871
Cw*0401	0.9273	0.8353	0.9636	0.9489	0.9818	0.9838