Large-scale protein-protein post-translational modification extraction with distant supervision and confidence calibrated BioBERT

Table 2 Train/test/val positive samples for each interaction type.

	Train		Val		Test		Total
	–	+	–	+	–	+	–	+
Acetylation	31	5	2	1	9	1	42	7
Dephosphorylation	167	28	36	10	33	6	236	44
Deubiquitination	4	2	0	0	0	0	4	2
Methylation	22	10	5	1	22	4	49	15
Phosphorylation	862	139	118	21	227	44	1207	204
Ubiquitination	30	5	5	1	5	1	40	7
Total	1116	189	166	34	296	56	1578	279

The negative samples count against a interaction type is solely to indicate how many were derived from abstracts describing the interaction type. All negative samples belong to a single class “other/Negative”

ISSN: 1471-2105