Table 2.

AUPRC over random of ETNA, MUNK, IsoRank, and HubAlign for predicting cross-species gene pairs that share GO annotations based on 5-fold cross validation.^a

Species pair	AUPRC (over random)
	ETNA	MUNK	IsoRank	HubAlign
H. sapiens $\to$M. musculus		0.590
H. sapiens $\leftarrow$M. musculus	0.805	0.452	0.561	0.459
H. sapiens $\to$S. cerevisiae		0.852
H. sapiens $\leftarrow$S. cerevisiae	1.390	0.927	0.775	0.726
H. sapiens $\to$D. melanogaster		0.607
H. sapiens $\leftarrow$D. melanogaster	0.724	0.650	0.594	0.532
H. sapiens $\to$C. elegans		0.515
H. sapiens $\leftarrow$C. elegans	0.572	0.444	0.221	0.370

^aAUPRC over random ${(\text{log}\hspace{0.17em}}_2\left(\frac{\text{AUPRC}}{\text{prior}}\right))$ is calculated to facilitate comparisons because each evaluation task has a different prior [proportion of positive examples in the gold standard (Section 2.3); the prior is different across organism pairs due to variability in the coverage of GO annotations within each organism, likely from differing levels of research and curation activity]. Thus, a score of 0 corresponds to random performance, and a score of 1 to a 2-fold improvement over random. For the four pairs of species (H. sapiens–M. musculus, H. sapiens–S. cerevisiae, H. sapiens–D. melanogaster, H. sapiens–C. elegans), the random priors are 0.059, 0.043, 0.044, and 0.044 respectively. Because MUNK’s predictions require choosing a source organism and a target organism, we present its performance for both directions (the arrow points from source to target). For each species pair, the top performance is shown in bold.