Skip to main content

Table 2 Prediction speed

From: P2Rank: machine learning based tool for rapid and accurate prediction of ligand binding sites from protein structure

Method Time
COACH (web server) 15 h (self reported estimate)
eFindSite (web server) \(6.9\pm 0\) h
COACH (stand-alone) \(6.4\pm 2\) h
GalaxySite (web server) 2 h (self reported estimate)
3DLigandSite (web server) 1–3 h (self reported estimate)
ISMBLab-LIG (web server) \(71\pm 2\) min
FTSite (web server) \(39\pm 3\) min
LISE (web server) \(39\pm 0.1\) min
MetaPocket 2.0 (web server) \(2.8\pm 0.4\) min
DeepSite (web server) \(38\pm 0.03\) s
SiteHound (stand-alone) \(12\pm 0.5\) s
P2Rank (stand-alone) \(6.8\pm 0.2\) s (cold start*)
  0.9 s (in larger dataset*)
Fpocket (stand-alone) \(0.2\pm 0.01\) s
  1. Average time required for LBS prediction on a single protein. Displayed is self reported estimate or a result of our test on a small dataset of 5 proteins á \(\sim \)2500 atoms. Stand-alone tools were tested on a single 3.7 GHz CPU core. For web servers the wall time from submitting a job to receiving the result was measured.
  2. *Difference is due to JVM initialization and model loading cost