DOCK Blaster:FAQ

DOCK Blaster is a free virtual screening service. These are the frequently asked questions. Questions that take more than a few lines deserve their own wiki article.

General

What is DOCK Blaster?

DOCK Blaster is a free web-based service for molecular docking and virtual screening. It uses UCSF DOCK 3.5.54 as a docking engine, and ZINC as a dockable database. It runs entirely on the computer cluster in the Shoichet Lab at Mission Bay in San Francisco. The graphical user interface comprises an expert system that is designed to help guide both novices and experts to useful results.

Is DOCK Blaster formally described in a paper? How to cite?

The performance of DOCK Blaster against benchmarking data is described in Irwin, Shoichet, Mysinger et al., "Automated docking screens: a feasibility study", J. Med. Chem (2009), in press. The documentation of the interface and the software that runs it is documented on this wiki.

Why DOCK Blaster? Docking programs have been around for ages. What's new?

We aim to serve a constituency of investigators who would like to use docking to discover new ligands for their targets, but find the barriers to setting up their own computational lab prevent them from doing so. Experts are still welcome to use DOCK Blaster, of course. But we are really aiming to serve biologists who seek new chemistry for their biological targets.

How do I use DOCK Blaster?

Please see What you need to start.

How much does DOCK Blaster cost to use?

DOCK Blaster is free. It is supported by the US National Institutes of Health via grant GM71896 (to JJI and BKS). We reserve the right to limit usage by intensive users.

I am having trouble using DOCK Blaster

If you began by uploaded a receptor file, aka "Start with a structure", by far the most common problem is with the formatting of the files. We recommend reading this and also looking at some examples and also sample data.

If you began with a PDB code, aka "Start with a PDB Code", by far the most common problem is that there is no single identifiable ligand in the file. If this is the case, you need to download the PDB file yourself, and, specifying the binding site, use the "Start with a structure" starting point.

If you used "I'm feeling lucky", and you are having problems, we regret that you were not more lucky. We suggest trying "I'm feeling lucky" a couple of more times, and perhaps your luck will change.

More detailed information about problems may be found here:

• Problems page for a general overview of all problems.

• Bugs page for information about bugs. Please report bugs to support at docking.org.

• DOCK Blaster:Problems page for DOCK Blaster specific problems.

Can DOCK Blaster be trusted?

Please see this.

I think DOCK Blaster has crashed

Please read this.

I have privacy concerns about my data

Please see our Privacy Policy

DOCK Blaster is reporting a problem with which I disagree

Please write to support at docking.org explaining the problem.

DOCK Blaster made completely wrong suggestions for my target

DOCK Blaster is a research tool that aims to help people discover new ligands. Unfortunately, docking can fail for many reasons, as documented by an extensive literature. Our benchmarking studies (Irwin et al, J Med Chem, 2009) suggest that black box docking works on perhaps 18% of eligible targets. Thus there is a rather high probability that your target is not among the successful ones. Finally, it is important to remain skeptical of any docking results, whether they are produced by an expert or a machine like DOCK Blaster.

Is there any evidence that DOCK Blaster works?

DOCK Blaster uses the same docking technology (DOCK 3.5.54) and database technology (ZINC) as is used in the Shoichet Lab for all their (publications). The DUD Paper (Huang, Shoichet Irwin, J Med Chem, 2006) assessed the performance of docking, some of it fully automated. Recently, the DOCK Blaster paper (Irwin et al, J Med Chem 2009, in press) assessed fully automated docking against over 7000 targets drawn from the PDB. The results indicate that automated docking screens produce viable-looking results perhaps 18% of the time. Importantly, the automatic self-assessment feature of DOCK Blaster allows you to tell when this is true.

What is your motivation to create and support DOCK Blaster?

We felt that many investigators who would like to use virtual screening were not using it because of several barriers to entry. We hope that DOCK Blaster will fill an niche and give access to this technology to investigators who previously were put off by the requirements to set up a docking laboratory.

Can I get my own private copy of DOCK Blaster?

You may request a copy of DOCK Blaster by writing to John Irwin at jji at cgl dot ucsf dot edu. We will be making the scripts available, but it will take us some time following publication (Sept 2009) to get them into a state we can release.

Does DOCK Blaster work better on some targets than on others?

Yes. Now, which ones? The paper (Irwin et al, J Med Chem, 2009, in press) documents 1,298 targets for which DOCK Blaster produces viable-looking results. You can find these on our website. TableS12.xls

How long does DOCK Blaster take?

Submitted to an unloaded cluster, preliminary docking usually runs in 30-60 minutes. A full database screen, which may or may not be warranted by the preliminary docking, can take hours to days, depending on the binding site and the database docked. If the cluster is loaded it will obviously take longer to deliver results. We aim to provide informative messages to manage your expectations. Thank you for your patience! More information here: DOCK Blaster:Timings

What do DOCK Blaster results mean?

Top scoring hits from a docking screen are computer-generated predictions of small molecules that may bind to your protein. Surveying the literature, typical prediction rates from docking screens vary wildly, but are often of a rate of only a few percent, although there are famous exceptions (Doman 2002). In a typical screen, purchasing 50 compounds selected from among the top 500, we would consider ourselves fortunate if 3 compounds showed micromolar affinity. What this means in practice is that docking hit lists should not be simply purchased willy-nilly. We recommend holding a hit picking party to select compounds for purchase and test based on a range of opinions. We would be astonished to see a confirmed hit rate over 10%, meaning that we expect at least 9/10 top scoring docking compounds WILL NOT BIND. Caveat emptor.

Can I used DOCK Blaster with a different docking program?

Currently, DOCK Blaster only supports DOCK 3.5.54. It has not escaped our notice that the overall system may be appropriate for other docking engines. If this option interests you, please write us, making the case for a particular docking program.

Can I used DOCK Blaster with my own database of small molecules?

Yes, to some extent. You can upload ligands to ZINC to create an user uploaded subset, 1000 molecules at a time. These subsets may be docked using DOCK Blaster.

A number of subsets of ZINC are available for docking. We may be able to create large custom subsets to suit your purposes. If you do not see what you need in the "by subset" page of ZINC, please write us at support at docking.org to discuss.

For most purposes, we recommend the use of the "fragment-like" or "lead-like" subsets of ZINC.

Why is it called "DOCK Blaster"?

The name is a tribute to BLAST, the famous sequence alignment and searching tool of NCBI. And of course DOCK. If you have a better name, let us know, there is a prize. ;-)

What is the history of DOCK Blaster?

• 1982-1993: DOCK Blaster draws heavily upon the manual docking protocol developed in Tack Kuntz's lab.
• 1989-2000 Further refinements of the manual protocol in Brian Shoichet's lab.
• 2000 Preliminary work by John Irwin
• 2001 First MakeDOCK script by David M. Lorber in Brian Shoichet's lab.
• 2001 Further refinements to MakeDOCK by Binqing Wei
• 2002 Refinements and new contributions by Austin N. Kirschner
• 2004-2006 Niu Huang's work on DUD was the first real test of semi-automated docking
• 2003-06 DOCK Blaster with GUI starts to be used by members of the Shoichet Lab.
• 2007-03 Alpha testing with 10+ users outside UCSF.
• 2007-09 DOCK Blaster can "do DUD automatically"
• 2007-10 Beta testing with 30+ users outside UCSF.
• 2008 Francesco Colizzi develops "be_blasti" and "auto_dock" enabling docking from a PDB code alone.
• 2009-01 Number of users > 50
• 2009 Eddie Cao performs extensive testing.
• 2009-05 Manuscript submitted
• 2009-08 Manuscript accepted, DOCK Blaster becomes public.
• 2009-09 Manuscript published.

I have a question that is not listed here...

Please use 'search' to check the documentation, use google, or try to ask our panel of experts in the blaster-fans@docking.org mailing list. If you still can't get an answer, please write support at docking.org.