Docking@Home

Docking@Home (also written Docking@home or D@H) was a volunteer computing project that used the BOINC (Berkeley Open Infrastructure for Network Computing) platform to study protein-ligand docking — the prediction of how a small molecule (ligand) binds to a protein target. The project was created by computer scientist and computational biologist Michela Taufer and her research group, first at the University of Texas at El Paso (UTEP) and later at the University of Delaware (UDel), where it was based for most of its operational life.

Docking@Home
Docking@Home BOINC screensaver, animating a ligand docking simulation in progress
The Docking@Home project homepage
Project
Status	Completed
Category	Biochemistry / Drug Discovery
Compute	CPU
Requires	None
Development
Developer	Michela Taufer and the Global Computing Lab
Author	Michela Taufer
Sponsor	University of Texas at El Paso; University of Delaware; National Science Foundation (NSF)
Initial release	December 1, 2005  (21 years ago)
Completed	May 23, 2014  (12 years ago)
Software
Written in	C, C++ (CHARMM)
Operating system	Linux, macOS, Windows
BOINC statistics
Stats as of	April 7, 2014  (12 years ago)
Total users	98,512
Total hosts	264,535
Metadata
Website	http://docking.cis.udel.edu/

The Docking@Home BOINC screensaver, showing a ligand docking simulation in progress. Video of the screensaver in motion.

Using the CHARMM (Chemistry at HARvard Macromolecular Mechanics) molecular simulation program, Docking@Home distributed an extensive conformational search of protein-ligand geometries to volunteers' home computers, with the ultimate goal of contributing to the discovery of new pharmaceutical drugs. The project was officially retired on May 23, 2014.

The Docking@Home project homepage, University of Delaware

History

The project's origins trace to the Global Computing Lab (GCLab), a research group led by Michela Taufer that began life at UTEP. Drawing on her earlier experience creating Predictor@home (the very first BOINC-based project, used to study protein structure prediction), Taufer brought to Docking@Home a similar emphasis on result validation through homogeneous redundancy — sending duplicate work units only to numerically identical machines so that results from independent volunteers could be compared bit-for-bit. The project's first volunteer newsletter was issued in December 2006, with project news posts referencing ongoing operations dating back to at least 2005.

In the fall of 2007, the Global Computing Lab and the Docking@Home project relocated from UTEP to the University of Delaware. The BOINC project news service announced the move to volunteers directly, advising that the server transition would require existing participants to detach and re-attach to the project once it was back online in Delaware. The relocated project subsequently launched a refreshed website at the University of Delaware in January 2009.

Throughout its operation the project's scientific direction was guided by collaborators across several institutions, including David P. Anderson (BOINC's creator, University of California, Berkeley), Charles L. Brooks III (The Scripps Research Institute), Patricia J. Teller (UTEP), and Roger S. Armen (Thomas Jefferson University). A companion outreach initiative, ExSciTecH (Expanding Volunteer Computing to Explore Science, Technology, and Health), was developed by the same research group to broaden participation in Docking@Home and related volunteer-computing efforts.

Retirement

On April 7, 2014, the Docking@Home team announced that the project would be retired due to a lack of resources to continue maintaining it. New jobs stopped being distributed on April 30, 2014, and the server stopped accepting results entirely on May 23, 2014, the date generally cited as the project's official retirement.

Over its roughly nine years of operation, Docking@Home had attracted 98,512 volunteers contributing computing power from 264,535 distinct hosts. The project granted a cumulative total of 5,422,290,917 BOINC credits, corresponding to an estimated 159,398,584 volunteer computing hours, or approximately 18,196 CPU-years of donated computation. The team made the project's full results dataset publicly available beginning in January 2014, ahead of the shutdown, and encouraged remaining volunteers to redirect their computers to other active volunteer computing projects.

Scientific goals and methods

The immediate scientific aim of Docking@Home was to refine a docking methodology based on CHARMM, which the project's developers considered one of the most accurate available methods for docking a flexible ligand against a rigid protein receptor. A secondary, longer-term goal was to apply these refined docking methods to specific biomedical problems of interest, including targets related to HIV.

Conformational search and clustering

A small-molecule ligand (sticks) positioned within the binding pocket of a protein receptor (surface), the kind of geometry Docking@Home's volunteers searched for.

Each docking calculation explores an enormous space of possible ligand positions, orientations, and internal conformations relative to a target protein's binding site. For a flexible ligand with $n$ rotatable internal bonds, combined with the six degrees of freedom describing its rigid-body position and orientation relative to the receptor, the size of the conformational space that must be sampled grows roughly as

$N\approx \prod _{i=1}^n\frac{2\pi }{\mathrm{\Delta }{\theta }_i}\times V_{\text{site}}$

where $\mathrm{\Delta }{\theta }_i$ is the angular sampling resolution for the $i$-th rotatable bond and $V_{\text{site}}$ represents the translational and rotational search volume of the binding site. This combinatorial growth is the basic reason docking benefits so strongly from being distributed across thousands of independent volunteer computers: each host can search a different region of the space in parallel, with results pooled afterward. Volunteers' computers performed molecular dynamics-based docking runs that sampled this conformational space, generating large sets of candidate ligand geometries that needed to be sorted into clusters of similar, low-energy poses before a "winning" (near-native) conformation could be identified.

As the project matured, its computer-science contributions increasingly focused on how to perform this clustering efficiently at scale. Researchers affiliated with the project, in particular Trilce Estrada and Boyu Zhang working with Taufer, developed MapReduce-based methods for classifying protein-ligand binding geometries and for capturing key scientific properties from the resulting datasets without having to physically move very large volumes of distributed data. Later work extended this clustering approach so it could run efficiently on large supercomputing systems, enabling scalable and accurate classification of the very large pools of ligand geometries the project's volunteers had generated.

Result validation

Like Predictor@home before it, Docking@Home relied on homogeneous redundancy to validate the numerically sensitive results returned by participants' machines, sending duplicate copies of a given work unit only to computers of matching architecture so that outputs could be compared for exact agreement rather than approximate similarity. The project's developers also built dedicated tools, such as the EmBOINC emulator, to study and improve how BOINC scheduling policies affected the throughput and reliability of volunteer projects like Docking@Home.

Community

Docking@Home built an active volunteer community supported by project message boards and a small team of forum moderators. The project received attention in the general press: in June 2009, Newswise published a feature encouraging computer owners to donate idle processing time to Docking@Home and similar disease-research projects.

The project's research team also studied the demographics of its own volunteer base. A 2013 study co-authored by Trilce Estrada and Michela Taufer specifically benchmarked gender differences among participants in volunteer computing projects, using Docking@Home as a case study.

Scientific publications

The following is a list of peer-reviewed papers and other scientific publications produced by the Docking@Home research team, as catalogued on the official BOINC Publications by Project page.

1. (2017).Enabling scalable and accurate clustering of distributed ligand geometries on supercomputers. Parallel Computing. DOI: 10.1016/j.parco.2017.02.005.
2. (2013).On Efficiently Capturing Scientific Properties in Distributed Big Data without Moving the Data: A Case Study in Distributed Structural Biology Using MapReduce. 2013 IEEE 16th International Conference on Computational Science and Engineering (CSE). DOI: 10.1109/CSE.2013.28.
3. (2013).Benchmarking Gender Differences in Volunteer Computing Projects. 2013 IEEE 9th International Conference on eScience (eScience). DOI: 10.1109/eScience.2013.29.
4. (2012).ExSciTecH: Expanding Volunteer Computing to Explore Science, Technology, and Health.
5. (2012).A scalable and accurate method for classifying protein-ligand binding geometries using a MapReduce approach. Computers in Biology and Medicine. DOI: 10.1016/j.compbiomed.2012.05.001.
6. (2012).Reengineering High-throughput Molecular Datasets for Scalable Clustering Using MapReduce. 2012 IEEE 14th Int'l Conf. on High Performance Computing and Communication (HPCC) & 2012 IEEE 9th Int'l Conf. on Embedded Software and Systems (ICESS). DOI: 10.1109/HPCC.2012.54.
7. (2011).Evaluation of Several Two-Step Scoring Functions Based on Linear Interaction Energy, Effective Ligand Size, and Empirical Pair Potentials for Prediction of Protein-Ligand Binding Geometry and Free Energy. Journal of Chemical Information and Modeling. DOI: 10.1021/ci1003009.
8. (2011).Providing Quality of Science in Volunteer Computing. Communication (HPCC). DOI: 10.1109/HPCC.2011.19.
9. (2010).Automatic selection of near-native protein-ligand conformations using a hierarchical clustering and volunteer computing. Proceedings of the First ACM International Conference on Bioinformatics and Computational Biology. DOI: 10.1145/1854776.1854807.
10. (2010).A fast protein-ligand docking algorithm based on hydrogen bond matching and surface shape complementarity. Journal of Molecular Modeling. DOI: 10.1007/s00894-009-0598-7.
11. (2009).Computational multiscale modeling in protein-ligand docking. IEEE Engineering in Medicine and Biology Magazine. DOI: 10.1109/MEMB.2009.931789.
12. (2009).Balancing Scientist Needs and Volunteer Preferences in Volunteer Computing Using Constraint Optimization. Computational Science – ICCS 2009.
13. (2009).Performance Prediction and Analysis of BOINC Projects: An Empirical Study with EmBOINC. Journal of Grid Computing. DOI: 10.1007/s10723-009-9126-3.
14. (2009).EmBOINC: An emulator for performance analysis of BOINC projects. Distributed Processing (IPDPS). DOI: 10.1109/IPDPS.2009.5161135.
15. (2009).Modeling Job Lifespan Delays in Volunteer Computing Projects. 2009 9th IEEE/ACM International Symposium on Cluster Computing and the Grid. DOI: 10.1109/CCGRID.2009.69.
16. (2008).A distributed evolutionary method to design scheduling policies for volunteer computing. the 2008 conference. DOI: 10.1145/1366230.1366282.
17. (2007).Evaluation of IEEE 754 floating-point arithmetic compliance across a wide range of heterogeneous computers. the 2007 conference. DOI: 10.1145/1347787.1347793.
18. (2007).Moving Volunteer Computing towards Knowledge-Constructed, Dynamically-Adaptive Modeling and Scheduling. 2007 IEEE International Parallel and Distributed Processing Symposium. DOI: 10.1109/IPDPS.2007.370668.
19. (2007).SimBA: A Discrete Event Simulator for Performance Prediction of Volunteer Computing Projects. 21st International Workshop on Principles of Advanced and Distributed Simulation. DOI: 10.1109/PADS.2007.27.
20. (2006).The Effectiveness of Threshold-Based Scheduling Policies in BOINC Projects. 2006 Second IEEE International Conference on e-Science and Grid Computing. DOI: 10.1109/E-SCIENCE.2006.261172.

See also

• BOINC
• Predictor@home
• World Community Grid
• Volunteer computing
• Protein-ligand docking
• CHARMM