Abstract: AutoDock Vina (Vina) is a widely adopted molecular docking tool, often regarded as a standard or used as a baseline in numerous studies. However, its computational process is highly time-consuming. The pioneering field-programmable gate array (FPGA)-based accelerator of Vina, known as Vina-FPGA, offers a high energy-efficiency approach to speed up the docking process. However, the computation modules in the Vina-FPGA design are not efficiently used. This is due to Vina exhibiting irregular behaviors in the form of nested loops with changing upper bounds and differing control flows. Fortunately, Vina employs the Monte Carlo iterative search method, which requires independent computations for different random initial inputs. This characteristic provides an opportunity to implement further parallel computation designs. To this end, this paper proposes Vina-FPGA2, an inter-module pipeline design for further accelerating Vina-FPGA. First, we use individual computational task (Task) independence by sequentially filling Tasks into computation modules. Then, we implement an inter-module pipeline parallel design by the Tag Checker module and architectural modifications, named Vina-FPGA2-Baseline. Next, to achieve resource-efficient hardware implementation, we describe it as an optimization problem and develop a reinforcement learning-based solver. Targeting the Xilinx UltraScale XCKU060 platform, this solver yields a more efficient implementation, named Vina-FPGA2-Enhanced. Finally, experiments show that Vina-FPGA2-Enhanced achieves an average 12.6× performance improvement over the central processing unit (CPU) and a 3.3× improvement over Vina-FPGA. Compared to Vina-GPU, Vina-FPGA2 achieves a 7.2× enhancement in energy efficiency.

Vina-FPGA2：基于模块间流水线的高层级并行硬件加速的分子对接工具

[1]Alhossary A, Handoko SD, Mu YG, et al., 2015. Fast, accurate, and reliable molecular docking with QuickVina 2. Bioinformatics, 31(13):2214-2216.

[2]Belletti F, Cotallo M, Cruz A, et al., 2009. Janus: an FPGA-based system for high-performance scientific computing. Comput Sci Eng, 11(1):48-58.

[3]Berman HM, Westbrook J, Feng ZK, et al., 2000. The Protein Data Bank. Nucleic Acids Res, 28(1):235-242.

[4]Caballero J, 2021. The latest automated docking technologies for novel drug discovery. Expert Opin Drug Discov, 16(6):625-645.

[5]Carvajal-Patiño JG, Mallet V, Becerra D, et al., 2025. RNAmigos2: accelerated structure-based RNA virtual screening with deep graph learning. Nat Commun, 16(1):2799.

[6]Chaudhary KK, Mishra N, 2016. A review on molecular docking: novel tool for drug discovery. JSM Chem, 4(43):1029.

[7]Chen YC, 2015. Beware of docking! Trends Pharmacol Sci, 36(2):78-95.

[8]Choi YK, Cong J, Fang ZM, et al., 2019. In-depth analysis on microarchitectures of modern heterogeneous CPU-FPGA platforms. ACM Trans Reconfig Technol Syst, 12(1):4.

[9]Corso G, Stärk H, Jing BW, et al., 2023. DiffDock: diffusion steps, twists, and turns for molecular docking. Proc 11^th Int Conf on Learning Representations, p.1-12.

[10]Ding J, Tang SD, Mei ZM, et al., 2023. Vina-GPU 2.0: further accelerating AutoDock Vina and its derivatives with graphics processing units. J Chem Inform Model, 63(7):1982-1998.

[11]Gaillard T, 2018. Evaluation of AutoDock and AutoDock Vina on the CASF-2013 benchmark. J Chem Inform Model, 58(8):1697-1706.

[12]Goodsell DS, Sanner MF, Olson AJ, et al., 2021. The AutoDock suite at 30. Protein Sci, 30(1):31-43.

[13]Gorgulla C, Boeszoermenyi A, Wang ZF, et al., 2020. An open-source drug discovery platform enables ultra-large virtual screens. Nature, 580(7805):663-668.

[14]Handoko SD, Ouyang X, Su CTT, et al., 2012. Quick-Vina: accelerating AutoDock Vina using gradient-based heuristics for global optimization. IEEE/ACM Trans Comput Biol Bioinform, 9(5):1266-1272.

[15]Hartshorn MJ, Verdonk ML, Chessari G, et al., 2007. Diverse, high-quality test set for the validation of protein-ligand docking performance. J Med Chem, 50(4):726-741.

[16]Hassan NM, Alhossary AA, Mu YG, et al., 2017. Protein-ligand blind docking using QuickVina-W with inter-process spatio-temporal integration. Sci Rep, 7(1):15451.

[17]He XH, Zhao LF, Tian YP, et al., 2024. Highly accurate carbohydrate-binding site prediction with DeepGlycanSite. Nat Commun, 15(1):5163.

[18]Hu YF, Zhu Y, Chen HY, et al., 2006. Communication latency aware low power NoC synthesis. Proc 43^rd Annual Design Automation Conf, p.574-579.

[19]Kirkpatrick S, Gelatt CD Jr, Vecchi MP, 1983. Optimization by simulated annealing. Science, 220(4598):671-680.

[20]Ling M, Lin QD, Chen RQ, et al., 2022. Vina-FPGA: a hardware-accelerated molecular docking tool with fixed-point quantization and low-level parallelism. IEEE Trans Very Large Scale Integr Syst, 31(4):484-497.

[21]Ling M, Feng ZH, Chen RQ, et al., 2024. Vina-FPGA-cluster: multi-FPGA based molecular docking tool with high-accuracy and multi-level parallelism. IEEE Trans Biomed Circ Syst, 18(6):1321-1337.

[22]McNutt AT, Francoeur P, Aggarwal R, et al., 2021. GNINA 1.0: molecular docking with deep learning. J Cheminform, 13(1):43.

[23]Michaud-Agrawal N, Denning EJ, Woolf TB, et al., 2011. MDAnalysis: a toolkit for the analysis of molecular dynamics simulations. J Comput Chem, 32(10):2319-2327.

[24]Mittal S, 2020. A survey of FPGA-based accelerators for convolutional neural networks. Neur Comput Appl, 32(4):1109-1139.

[25]Muhammed MT, Aki-Yalcin E, 2024. Molecular docking: principles, advances, and its applications in drug discovery. Lett Drug Des Discov, 21(3):480-495.

[26]Pechan I, Feher B, 2011. Molecular docking on FPGA and GPU platforms. Proc 21^st Int Conf on Field Programmable Logic and Applications, p.474-477.

[27]Pechan I, Fehér B, Bérces A, 2010. FPGA-based acceleration of the AutoDock molecular docking software. Proc 6^th Conf on Ph.D. Research in Microelectronics & Electronics, p.1-4.

[28]Salmaso V, Moro S, 2018. Bridging molecular docking to molecular dynamics in exploring ligand-protein recognition process: an overview. Front Pharmacol, 9:923.

[29]Santos-Martins D, Solis-Vasquez L, Tillack AF, et al., 2021. Accelerating AutoDock4 with GPUs and gradient-based local search. J Chem Theory Comput, 17(2):1060-1073.

[30]Shawahna A, Sait SM, El-Maleh A, 2019. FPGA-based accelerators of deep learning networks for learning and classification: a review. IEEE Access, 7:7823-7859.

[31]Solis-Vasquez L, Koch A, 2017. A performance and energy evaluation of OpenCL-accelerated molecular docking. Proc 5^th Int Workshop on OpenCL, p.1-11.

[32]Solis-Vasquez L, Koch A, 2018. A case study in using OpenCL on FPGAs: creating an open-source accelerator of the AutoDock molecular docking software. Proc 5^th Int Workshop on FPGAs for Software Programmers, p.1-10.

[33]Solis-Vasquez L, Santos-Martins D, Tillack AF, et al., 2020. Parallelizing irregular computations for molecular docking. Proc 10^th IEEE/ACM Workshop on Irregular Applications: Architectures and Algorithms (IA3), p.12-21.

[34]Su M, Yang Q, Du Y, et al., 2018. Comparative assessment of scoring functions: the CASF-2016 update. J Chem Inform Model, 59(2):895-913.

[35]Tang S, Chen R, Lin M, et al., 2022. Accelerating AutoDock Vina with GPUs. Molecules, 27(9):3041.

[36]Trott O, Olson AJ, 2010. AutoDock Vina: improving the speed and accuracy of docking with a new scoring function, efficient optimization, and multithreading. J Comput Chem, 31(2):455-461.

[37]Vittorio S, Lunghini F, Morerio P, et al., 2024. Addressing docking pose selection with structure-based deep learning: recent advances, challenges and opportunities. Comput Struct Biotechnol J, 23:2141-2151.

[38]Xilinx, 2023. Xilinx Power Estimator (XPE). https://www.xilinx.com/products/technology/power/xpe.html [Accessed on Oct. 22, 2024].

[39]Zeng S, Liu J, Dai G, et al., 2024. FlightLLM: efficient large language model inference with a complete mapping flow on FPGAs. Proc ACM/SIGDA Int Symp on Field Programmable Gate Arrays, p.223-234.

[40]Zhou X, Ling M, Lin Q, et al., 2023. Effectiveness analysis of multiple initial states simulated annealing algorithm, a case study on the molecular docking tool AutoDock Vina. IEEE/ACM Trans Comput Biol Bioinform, 20(6):3830-3841.