TABLE 2.

Pairwise comparison between the generalization performance of models trained using only the mutation data^a

	Model A	Model B	Mean A	Mean B	P(A > B)	P(A == B)	P(A < B)
Clade 20C	LANDMark	Extra Trees	0.965 ± 0.035	0.968 ± 0.037	0.001	0.996	0.003
	LANDMark	Logistic Regression	0.965 ± 0.035	0.972 ± 0.035	0.001	0.990	0.009
	LANDMark	K-Nearest Neighbors	0.965 ± 0.035	0.875 ± 0.051	0.948	0.052	0
	LANDMark	Linear SVC	0.965 ± 0.035	0.973 ± 0.030	0	0.999	0.001
	Extra Trees	Logistic Regression	0.968 ± 0.037	0.972 ± 0.035	0.004	0.985	0.011
	Extra Trees	K-Nearest Neighbors	0.968 ± 0.037	0.875 ± 0.051	0.969	0.031	0
	Extra Trees	Linear SVC	0.968 ± 0.037	0.973 ± 0.030	0.003	0.985	0.012
	Logistic Regression	K-Nearest Neighbors	0.972 ± 0.035	0.875 ± 0.051	0.952	0.048	0
	Logistic Regression	Linear SVC	0.972 ± 0.035	0.973 ± 0.030	0	0.999	0.001
	K-Nearest Neighbors	Linear SVC	0.875 ± 0.051	0.973 ± 0.030	0	0.023	0.977
Clade 21J	LANDMark	Extra Trees	0.892 ± 0.044	0.896 ± 0.042	0	1	0
	LANDMark	Logistic Regression	0.892 ± 0.044	0.886 ± 0.045	0.002	0.998	0
	LANDMark	K-Nearest Neighbors	0.892 ± 0.044	0.764 ± 0.049	1	0	0
	LANDMark	Linear SVC	0.892 ± 0.044	0.882 ± 0.042	0.008	0.991	0
	Extra Trees	Logistic Regression	0.896 ± 0.042	0.886 ± 0.045	0.009	0.990	0
	Extra Trees	K-Nearest Neighbors	0.896 ± 0.042	0.764 ± 0.049	1	0	0
	Extra Trees	Linear SVC	0.896 ± 0.042	0.882 ± 0.042	0.023	0.976	0
	Logistic Regression	K-Nearest Neighbors	0.886 ± 0.045	0.764 ± 0.049	0.999	0.001	0
	Logistic Regression	Linear SVC	0.886 ± 0.045	0.882 ± 0.042	0.004	0.994	0.001
	K-Nearest Neighbors	Linear SVC	0.764 ± 0.049	0.882 ± 0.042	0	0.003	0.997
All Clades	LANDMark	Extra Trees	0.982 ± 0.007	0.982 ± 0.006	0	1	0
	LANDMark	Logistic Regression	0.982 ± 0.007	0.984 ± 0.005	0	1	0
	LANDMark	K-Nearest Neighbors	0.982 ± 0.007	0.950 ± 0.018	0.035	0.965	0
	LANDMark	Linear SVC	0.982 ± 0.007	0.985 ± 0.018	0.0	1	0.0
	Extra Trees	Logistic Regression	0.982 ± 0.006	0.984 ± 0.005	0.0	1	0.0
	Extra Trees	K-Nearest Neighbors	0.982 ± 0.006	0.950 ± 0.018	0.027	0.973	0
	Extra Trees	Linear SVC	0.982 ± 0.006	0.985 ± 0.018	0	1	0
	Logistic Regression	K-Nearest Neighbors	0.984 ± 0.005	0.950 ± 0.018	0.043	0.957	0
	Logistic Regression	Linear SVC	0.984 ± 0.005	0.985 ± 0.018	0	1	0
	K-Nearest Neighbors	Linear SVC	0.950 ± 0.018	0.985 ± 0.018	0	0.947	0.053

^a A Bayesian t-test was used to determine the probability that the balanced accuracy score of model A either exceeds, is lower, or is equivalent to the performance of model B. Feature selection using Triglav was not performed to generate these results.