TBGA: a large-scale Gene-Disease Association dataset for Biomedical Relation Extraction

BMC Bioinformatics

Table 4 RE models performance on TBGA dataset

Model	Strategy	AUPRC	P@50	P@100	P@250	P@500	P@1000
CNN	AVE	0.422	0.780	0.760	0.744	0.696	0.625
CNN	ATT	0.403	0.780	0.760	0.788	0.710	0.624
PCNN	AVE	0.426	0.780	0.780	0.744	0.720	0.664
PCNN	ATT	0.404	0.760	0.750	0.744	0.700	0.628
BiGRU	AVE	0.437	0.620	0.720	0.724	0.730	0.678
BiGRU	ATT	0.423	0.760	0.750	0.748	0.726	0.666
BiGRU-ATT	AVE	0.419	0.740	0.740	0.748	0.694	0.615
BiGRU-ATT	ATT	0.390	0.680	0.760	0.756	0.702	0.631
BERE	AVE	0.419	0.700	0.710	0.720	0.704	0.620
BERE	ATT	0.445	0.780	0.780	0.800	0.764	0.709

Columns represent, from left to right, the considered RE model, the aggregation strategy, the AUPRC score, as well as the P@50, P@100, P@250, P@500, and P@1000 scores. For each measure, bold values represent the best scores

ISSN: 1471-2105