Home   |   Contact
Computing Speed

.
The speed of ligand docking calculations depends on the settings of genetic algorithm search. Currently, Lead Finder provides two types of calculation regime: a more accurate and exhaustive search in the default docking regime and a faster search at the cost of slight decrease of accuracy in the screening regime. While the setting for the default docking settings have been adjusted to achieve maximum docking accuracy, the settings for the screening regime were fine-tuned to achieve maximum speed of calculations at the cost of slight (~5%) decrease in Docking Success Rate.
To benchmark the speed of docking calculations, a set of 407 diverse protein-ligand complexes (used also for Docking Success Rate measurements) was used. All calculations were performed in two regimes: docking and screening. The above figure presents results of the current benchmarking study with average times needed to dock a ligand as a function of ligand's degrees of freedom (number of freely rotatable bonds). The screening regime is 2-4 times faster than the docking regime. In screening regime, Lead Finder docks a ligand with 6 freely rotatable bonds within 30 sec on average.


This chart illustrates how the speed of docking calculations depends on the number of ligand's freely rotatable bonds in two regimes (docking and screening). Each mark represents the average processing time for compounds from the docking success rate benchmarking set of 407 complexes.

Protein target

Time pre liganda, seconds

Total screening timeb, hours

Volume of the energy gridc, Å3

Monoamine oxydase A

9.00

7.50

4300

HIV-1 reverse transcriptase

10.4

8.70

5580

Thyroid hormone receptor beta1

10.6

9.19<

6200

Monoamine oxydase B

11.0

8.87

6720

Vitamin D receptor

13.4

11.18

8350

Dihydroorotate dehydrogenase

14.7

12.24

9730

Dihydrofolate reductase

15.0

12.48

9400

HMG-CoA reductase

15.2

12.67

9390

Leukotriene A4 hydrolase

15.8

13.16

9480

PPAR-γ

18.4

15.36

14670

a) Time pre ligand indicates average time (in seconds) of docking one compound from the library (on the single processor core).
b) Total screening time  total time (in hours) of virtual library screening on the computer cluster.
c) Volume of the energy grid  indicates the volume of the energy grid (for more details on grid definition see Docking algorithm section).
We also benchmarked the speed of ligand docking in a number of real-life virtual screening studies. In one study, a commercial library of 300,000 compounds (STK library by Vitas-M Laboratory) was screened against a number of protein targets. Using a cluster of 16 computational nodes (dual Xeon 5150 2.6 GHz) we were able to screen the entire library in 12 to 24 hours depending on the protein target. The average time for docking of a compound from the library for a particular protein is provided in the table above.
To estimate relative speed of docking calculations in different computing environments, the following table can be used as a guide:
Time needed to screen a library of 300,000 compounds against different protein targets on a 16-node (dual Xeon 5150 2.6 GHz) computer cluster.

OS

Platform

CPU

Relative speed

Windows

x86_32

Xeon 5150 2.67GHz

1.7

Linux

x86_64

Core2Duo 2.4GHz

1.8

Linux

x86_64

Opteron 246 2GHz

1.0

Linux

x86_32

Athlon MP 1.67GHz

0.4



Relative speed of ligand docking calculations in different computing environments.
E-mail: info@biomoltech.com |  Phone: +1(416)238-1263 |  Fax: +1(416)352-6117
Mailing address: 226 York Mills Rd, Toronto, Ontario M2L 1L1, Canada
Privacy Policy | Terms of Use
© 2017 BioMolTech Corp.