Multi-Server Approach for High-Throughput Molecular Descriptors Calculation based on Multi-Linear Algebraic Maps
| Autor: | Reisel González-Pérez, Tatiana Avdeenko, Liesner Acevedo-Martínez, Longendri Aguilera-Mendoza, Yovani Marrero-Ponce, César R. García-Jacas, Stephen J. Barigye |
|---|---|
| Rok vydání: | 2014 |
| Předmět: |
Communication protocol
Theoretical computer science Workstation Computational complexity theory Computer science Performance Computation Quantitative structure activity relation TOMOCOMD-CARDD Article law.invention Platform of distributed tasks Software Chemical structure Structural Biology law Drug Discovery Theoretical model Multi-server architecture Throughput (business) Priority journal Distributed computing system business.industry Organic Chemistry Local area network Calculation Usability Models Theoretical 3D N-linear algebraic descriptors T-arenal Computer Science Applications Algorithm Task (computing) Computer engineering Molecular Medicine business QuBiLS-MIDAS Mathematics |
| Zdroj: | Molecular Informatics. 34:60-69 |
| ISSN: | 1868-1743 |
| DOI: | 10.1002/minf.201400086 |
| Popis: | The present report introduces a novel module of the QuBiLS-MIDAS software for the distributed computation of the 3D Multi-Linear algebraic molecular indices. The main motivation for developing this module is to deal with the computational complexity experienced during the calculation of the descriptors over large datasets. To accomplish this task, a multi-server computing platform named Tarenal was developed, which is suited for institutions with many workstations interconnected through a local network and without resources particularly destined for computation tasks. This new system was deployed in 337 workstations and it was perfectly integrated with the QuBiLSMIDAS software. To illustrate the usability of the T-arenal platform, performance tests over a dataset comprised of 15000 compounds are carried out, yielding a 52 and 60 fold reduction in the sequential processing time for the 2-Linear and 3-Linear indices, respectively. Therefore, it can be stated that the T-arenal based distribution of computation tasks constitutes a suitable strategy for performing high-throughput calculations of 3D Multi-Linear descriptors over thousands of chemical structures for posterior QSAR and/or ADME-Tox studies. © 2015 Wiley-VCH Verlag GmbH & Co. KGaA. |
| Databáze: | OpenAIRE |
| Externí odkaz: |
|
Plný text