Table 2 Inference speed comparison on SwissProt E-RXN ASA test set between EasIFA and the baseline algorithmsa
From: Multi-modal deep learning enables efficient and accurate annotation of enzymatic active sites
Methods | GPU/CPU | Knowledge base size | Number test set samples | Inference Time | Time pre sample |
|---|---|---|---|---|---|
EasIFA-NGb | 1 RTX3060 GPU | SwissProt E-RXN ASA dataset: 44,341sequences | 892 | 113 s | 0.127 s |
EasIFA-ESM | 129 s | 0.144 s | |||
EasIFA-SaProt | 146 s | 0.164 s | |||
BLASTp | CPU 1 threads | 225 s | 0.252 s | ||
CPU 16 threads | 131 s | 0.146 s | |||
CPU 1 threads | SwissProt: 569,516 sequences | 1212 s | 1.359 s | ||
CPU 16 threads | 262 s | 0.294 s | |||
AEGAN | RTX3060 1GPU + CPU 16 threads | 16,841 PDB | >48 h | >200 s | |
Schrodinger-SiteMap | CPU 16 threads | na | >24 h | >100 s |
