Automated Training for Algorithms That Learn from Genomic Data

Table 1

Cardinalities of the positive interim training sets for the 17 apicomplexan species gathered by ApicoAP-CS.

Apicomplexan species	Ortho-MCL^a	BLAST^b	Confirmed^c	All combined^d	Conflicts removed^e	Non-SP filtered^f
B. bovis	46	45	4	61	59	18
B. microti	51	50	0	61	58	23
C. hominis	17	24	0	28	25	1
C. muris	19	27	0	32	29	0
C. parvum	17	24	0	29	26	2
E. tenella	82	61	1	89	84	30
N. caninum	78	68	0	82	77	21
P. berghei	72	73	0	77	73	49
P. chabaudi	72	73	0	77	72	51
P. cynomolgi	70	72	0	77	73	31
P. falciparum	45	60	40	89	85	52
P. knowlesi	72	72	0	77	73	49
P. vivax	69	72	0	75	71	51
P. yoelii	70	68	3	77	73	41
T. annulata	45	47	0	56	54	23
T. parva	49	42	0	59	57	25
T. gondii	53	59	45	102	96	42

 ^aCardinality of the set gathered by ortholog search using OrthoMCL.
^bCardinality of the set gathered by ortholog search using the BLAST-based algorithm.
^cCardinality of the set containing experimentally confirmed positive/negative proteins.
^dCardinality of the set that is the union of the sets presented in column 2, 3 and 4.
^eCardinality of the union set when conflicts with the negative/positive set is removed.
^fCardinality of the final training set after proteins without signal peptides have been removed.