Table III. Variable prediction cutoffs.

If the cutoff values for prediction are increased to mean − sd, mean, mean + sd, and mean + 2 sd, the precision and recall values change as shown in the rows Medium to Highest from bottom to top. Hence, predictions can be chosen according to the precision and recall desired. The TP_pred. and FP_pred. are categorized according to their D_max scores. The highest D_max score for TP_pred. of a TP_gene is considered as the score for that gene and is used to categorize the TP_genes in the score classes.

Score Class	TPpred.	TPpred.	FPpred.	FPpred.	TPgenes	TPgenes	Recall	Precision	F Value
		%		%		%
Arabidopsis
Highest	1,100	3.98	772	1.38	1,076	5.84	0.05	0.59	0.09
Very high	3,738	13.53	2,974	5.32	3,497	18.99	0.23	0.56	0.33
High	8,723	31.57	13,160	23.52	7,073	38.4	0.58	0.45	0.51
Medium	12,075	43.7	31,944	57.09	6,168	33.49	0.89	0.34	0.49
Low	1,993	7.21	7,102	12.69	604	3.28	0.92	0.33	0.49
Total	27,629	100	55,952	100	18,418	100	0.92	0.33	0.49
Rice
Highest	3,651	11.18	2,846	2.75	3,482	16.57	0.15	0.56	0.24
Very high	6,439	19.72	7,202	6.96	5,609	26.69	0.39	0.5	0.44
High	10,491	32.13	25,350	24.5	7,379	35.11	0.71	0.37	0.48
Medium	9,872	30.23	54,706	52.87	4,125	19.63	0.89	0.25	0.39
Low	2,198	6.73	13,365	12.92	424	2.02	0.91	0.24	0.38
Total	32,651	100	103,469	100	21,019	100	0.91	0.24	0.38