CoMAGC: a corpus with multi-faceted annotations of gene-cancer relations

BMC Bioinformatics

Table 3 Inference rules for gene classification

Rule #	CGE	CCS	IGE	PT	Gene class
1	increased	normal- >cancer	*	causality	oncogene
2	decreased	cancer- >normal	unidentifiable	causality	oncogene
3	decreased	cancer- >normal	up-regulated	*	oncogene
4	decreased	normal- >cancer	*	causality	tumor suppressor gene
5	increased	cancer- >normal	unidentifiable	causality	tumor suppressor gene
6	increased	cancer- >normal	down-regulated	*	tumor suppressor gene
7	*	normal- >cancer	*	observation	biomarker
8	*	cancer- >normal	unidentifiable	observation	biomarker
9	decreased	cancer- >cancer	up-regulated	observation	biomarker
10	increased	cancer- >cancer	down-regulated	observation	biomarker

The rules are not symmetric to each other. For instance, Rule 4 states that a gene can be classified as a 'tumor suppressor gene’ when a decreased expression level of the gene causes cancer progression, regardless of the IGE value. Rule 2 is about a similar case, where a decreased expression level of a gene causes cancer regression. However, Rule 2 requires IGE to be 'unidentifiable’ to infer the gene as an 'oncogene’. Rule 3, not Rule 2, covers the case where CGE is 'up-regulated’ and PT is 'causality’. Also, Rule 2 does not cover the cases where IGE is 'down-regulated’ or 'unchanged’ since we expect such cases to be rarely reported in biology articles (cf. Section on Inference rule development). Rules 5 and 8 are designed in a similar way. The asterisk denotes all the pre-defined values for the corresponding concept.

ISSN: 1471-2105