Table 3.

Performance on disease–gene predictions: our BioEmbedS approach’s 10-fold CV-based result is compared against existing methods’ reported results on the EU-ADR disease–gene corpus

Model	Precision (%)	Recall (%)	F1-score (%)
BioEmbedS	77.13 ± 1.41	96.84 ± 2.67	85.84 ± 1.26
BioBERT v1.0	81.05	93.90	86.51
BioBERT v1.1	77.86	83.55	79.74
BeFree	75.10	97.70	84.6
Joint Ensemble learning	76.43	98.01	85.34

Note: BioBERT results are obtained from Lee et al. (2020); results on BeFree and Joint Ensemble learning models are obtained from Bravo et al. (2015) and Bhasuran and Natarajan (2018), respectively.

Bold-faced values indicate the best performance achieved in the corresponding column.