Semi-supervised learning for the identification of syn-expressed genes from fused microarray and in situ image data

Table 1 Selection of Constraints.

		Proportion of terms with lower p-values
λ ₊	λ _-	# stages ≥ 3	≥4
0.5	-	51%	48%
1.0	-	60%	56%
1.5	-	57%	49%
2.0	-	43%	46%
0.5	0.5	49%	44%
1.0	1.0	49%	52%
1.5	1.5	40%	59%
2.0	2.0	43%	47%

We show the proportion of ImaGO terms with lower p-values in cMoG compared to MoG for constraints from at least 3 or 4 stages, distinct constraint weights λ₊ and λ_- using positive constraints only respectively positive and negative constraints. Values exceeding 50% indicate an advantage of cMoG.

ISSN: 1471-2105