A graph-search framework for associating gene identifiers with documents

BMC Bioinformatics

Table 5 Features Used for Ranking

TestFile $\underset{\underset{}{\to}}{hasTerm}$	EntityType $\underset{\underset{}{\to}}{h a s S p a n}$	String $\underset{\underset{}{\to}}{h a s P o s s i b l e P r o t e i n^{- 1}}$
GeneSyn $\underset{\underset{}{\to}}{{synonym}^{- 1}}$	Labels $\underset{\underset{}{\to}}{hasPossibleProtein}$	String $\underset{\underset{}{\to}}{hasTerm}$
TestFile $\underset{\underset{}{\to}}{possibleProtNER}$	GeneId $\underset{\underset{}{\to}}{{hasGene}^{- 1}}$	TERM $\underset{\underset{}{\to}}{inFile}$
TrainFile $\underset{\underset{}{\to}}{hasGene}$	Labels $\underset{\underset{}{\to}}{annotates}$	Labels $\underset{\underset{}{\to}}{hasLikelyProtein}$
TrainFile $\underset{\underset{}{\to}}{{annotates}^{- 1}}$	String $\underset{\underset{}{\to}}{{hasLikelyProtein}^{- 1}}$	TestFile $\underset{\underset{}{\to}}{likelyProtNER}$
TestFile $\underset{\underset{}{\to}}{{annotates}^{- 1}}$	String $\underset{\underset{}{\to}}{likelyProt2syn}$	String $\underset{\underset{}{\to}}{possibleProt2syn}$
GeneSyn $\underset{\underset{}{\to}}{hasTerm}$	TrainFile $\underset{\underset{}{\to}}{hasTerm}$	GeneId $\underset{\underset{}{\to}}{synonym}$
String $\underset{\underset{}{\to}}{{hasSpan}^{- 1}}$	Labels $\underset{\underset{}{\to}}{hasSpanType}$	EntityType $\underset{\underset{}{\to}}{{hasSpanType}^{- 1}}$

The final set of features used by learning system, in the format $" t \overset{ℓ}{\to}'$ , where t is a node type, and ℓ is an edge label. The edge labels likelyProtNER and possibleProtNER are shortcut representing the baseline NER methods. The edge labels likelyProt2syn and possibleProt2syn are links between extracted protein names and softTFIDF-similar protein synonyms.

ISSN: 1471-2105