ISSN 0095-4527, Cytology and Genetics, 2012, Vol. 46, No. 3, pp. 172–179. © Allerton Press, Inc., 2012. Original Russian Text © O.M. Demchuk, P.A. Karpov, Ya.B. Blume, 2012, published in Tsitologiya i Genetika, 2012, Vol. 46, No. 3, pp. 55–64. 172 INTRODUCTION Modeling of intermolecular interactions is one of the most required computational tasks in structural bioinformatics [1]. Traditionally, computational plat- forms based on the use of a single central processing unit (CPU) or a multi-core computing environment (cluster solutions, grid) are used for calculations of this type of resource-intensive tasks in silico [2, 3]. Recently, calculation technologies using graphics pro- cessing units (GPUs) [4] and, specifically, CUDA (Compute Unified Device Architecture) have appeared. As a result, GPU calculations have been applied for solving a wide spectrum of scientific tasks [3], including calculations in the area of bioinformat- ics and molecular biology [5]. It positively affected the speed of certain computing operations and has led to a significant reduction in calculation time, necessary for performing a number of algorithms and bioinformat- ics research in general. The application of a GPU reduces the calculation time of DNA and protein sequence alignment 10 [6, 7] to 100 [8] times compared to calculations made using a central processor unit (CPU). It is shown that a GPU speeds up calculations for molecular dynamics tasks 10 to 100 times [9, 10], up to 130 times for quan- tum chemistry [11], and more than 200 times for wavelet analysis of mass spectrometry data [12]. Appli- cation of a GPU for the computation of an available protein surface reduces the calculation time 100 to 300 times, depending on the size of the modeled system [13]. It is not surprising that the quantity of software applications supporting the opportunity of molecular calculations using GPUs has increased rapidly [14]. The undisputable leader in the area of GPU calcula- tion technology developments is the nVidia company, and we particularly note its technology CUDA [15]. Almost a full list of applications supporting CUDA is presented on the webpage CUDA Zone of the official nVidia website (http://www.nvidia.com/). Presently, the list consists of more than 20 applications aimed at solving bioinformatics and structural biology tasks. Solutions supporting CUDA are presented for one of the most demanded instruments in bioinformat- ics—BLAST (The Basic Local Alignment Search Tool, NCBI-BLAST www.ncbi.nlm.nih.gov). These solutions are GPU-BLAST (http://eudoxus.cheme. cmu.edu/gpublast/), which works 4 times faster with search results identical to traditional BLASTn [16], and CUDA-BLASTP [17] aimed at scanning big arrays of aminoacid sequences. Also, these CUDA- supporting applications include software for convert- ing sequences in a format optimized for use in BLAST applications adapted for CUDA (http://www.nvidia. com/object/blastp_on_tesla.html) [17]. Another application—CUDASW++ (http:// sourceforge.net/projects/cudasw/files/)—is designed for conducting search for aminoacid sequences in databases using the Smith–Waterman algorithm [18, 19]. It is shown that using NVIDIA GeForce graphics accelerators (GTX 280 and GTX 295) the search time with the application of the CUDASW++ software reduces 10 to 50 times compared to NCBI-BLAST [18]. Based on the MEME (V. 3.5.4) [20] software package, a CUDA version—CUDA-MEME (mCUDA-MEME) (http://www.nvidia.com/object/meme_on_tesla.html) [21, 17]—was created for tasks, such as search for conservative motifs of sequences. Docking Small Ligands to Molecule of the Plant FtsZ Protein: Application of the CUDA Technology for Faster Computations O. M. Demchuk, P. A. Karpov, and Ya. B. Blume Institute of Food Biotechnology and Genomics, National Academy of Sciences of Ukraine, Kiev, 04123 Ukraine e-mail: demom79@gmail.com Received October 26, 2011 Abstract—The opportunities to apply the CUDA technology for faster computations in structural biology and bioinformatics are reviewed and analyzed. Using HEX 6.1 software, we performed a comparative analysis of the efficiency and the increase in quality after CUDA application. The work was conducted on the example of rigid docking of low-molecular-weight compounds of different classes on the surface of the Arabidopsis thaliana FtsZ. Several potential binding sites of benzimidazoles to the plant FtsZ were identified. DOI: 10.3103/S0095452712030048