Fragment-based docking: Development of the CHARMMing Web user interface as a platform for Computer-Aided Drug Design

Yuri Pevzner , Emilie Frugier , Vinushka Schalk , Amedeo Caflisch , and H. Lee Woodcock
J. Chem. Inf. Model., Just Accepted Manuscript

Abstract
Web-based front end interfaces to scientific applications are important tools that allow researchers to utilize a broad range of software packages with just an Internet connection and a browser. One such interface, CHARMMing (CHARMM interface and graphics), allows researcher to take advantage of the functionality of the powerful and widely used molecular software package CHARMM. CHARMMing incorporates tasks such as molecular structure analysis, dynamics, multi-scale modeling, and other techniques commonly used by computational life scientists. We have extended CHARMMing's capabilities to include a fragment-based docking protocol that allows users to perform molecular docking and virtual screening calculations either directly on the CHARMMing Web server or on local HPC resources using the self-contained job scripts generated via the Web interface. The docking protocol was evaluated by performing a series of "re-docking's" with direct comparison to top commercial docking software. Results of this evaluation showed that CHARMMing's docking implementation is comparable to many widely used software packages and validates the use of the new CHARMM generalized force field for docking and virtual screening.

SP
Make a directory and include the following files:

• make_dirs
  
• copy_files2
  
• glide-dock_SP_0.in
  

Example of SP.in file:
USECOMPMAE YES
NREPORT 30000
RINGCONFCUT 2.500000
GRIDFILE /home/eva/Zinc_screening/SP_1MV9/1MV9_no_wat_glide-grid.zip
LIGANDFILE /home/ZINC/druglike-Zinc-library-mol2/3_p0.0.mol2
LIGFORMAT mol2

1. Run the make_dirs script --> creates 88 folders (0-87)
   
2. Run the copy_files2 script --> leave the glide-dock_SP_0.in file out of its folder to run the script and when it is finished then cut+paste it to folder_0.
   
3. Use the submit_SP_all script to run the SP on apple cluster. In my case I used to put up to 12 jobs in each node ( modify this line of the script: for (( i = 76 ; i<=87; i++   )) ). Due to the fact that the jobs were put randomly to the nodes and in order to have up to 12 jobs on each node, I was using first some "fake" gromacs runs to use 12 cores and left the other 12 free for screening runs (Evi's idea! Thank's Evi!!).
4. When the run is finished you should check all the 88 directories and see if you have outputs like: e.g. glide-dock_SP_8_pv.maegz. If something goes wrong and the run has stopped then you will see e.g. glide-dock_SP_8_raw.maegz as an output. Then you should copy your glide-dock_SP_8.in to glide-dock_SP_8_b.in and run it from the ligand that it had stopped. Finally, you will have: glide-dock_SP_8_raw.maegz and glide-dock_SP_8_b_pv.maegz. Both are useful! Don't delete the first one!
5. In order to finish the SP process you have to take the top-scored 40.000 compounds from all the directories and put them to a new file (SP_top40000_1MV9_pv.maegz) to use as an input for the XP process.

You can run the GlideSortScript_ZINC.csh script to acheive this result. The content of the script is the following: (You have to add the possible raw.maegz files,too.)
/opt/schrodinger/utilities/glide_sort -n 40000 -o SP_top40000_1MV9_pv.maegz -r  SP_top40000_1MV9.rept -hbond_cut 0.00 -cvdw_cut 0.00 -metal_cut 10.00 \ db3_p0.0/glide-dock_SP_0_pv.maegz \db3_p0.1/glide-dock_SP_1_pv.maegz \db3_p0.2/glide-dock_SP_2_pv.maegz \db3_p0.3/glide-dock_SP_3_raw.maegz \db3_p0.3/glide-dock_SP_3_b_pv.maegz \ etc.

XP
Each XP directory includes:

• grid.zip file
  
• SP_top40000_1MV9_pv.maegz (output of SP)
  
• XP.in files (I split them into 8 files in my case) 

Example of XP.in file:
WRITEREPT YES
WRITE_RES_INTERACTION YES
WRITE_XP_DESC YES
USECOMPMAE YES
POSTDOCK_NPOSE 10
LIGAND_END 10000
LIGAND_START 5001 (this line is not included in the XP_1.in file as it starts from the beginning)
MAXREF 800RINGCONFCUT 2.500000
GRIDFILE /home/eva/XP_Zinc/1MV9_XP_Zinc_2_8/1MV9_no_wat_glide-grid.zip
LIGANDFILE /home/eva/XP_Zinc/1MV9_XP_Zinc_2_8/SP_top40000_1MV9_pv.maegz
PRECISION XP

So, I split it to 8 XP directories:
0 - 5000, 5001  - 10000, 10001 - 15000, ..., 35001 - 40000

Run each one on the cluster or on our PCs manually :
/opt/schrodinger/glide 1MVC_Zinc_XP1.in -HOST xgrid-server
/opt/schrodinger/glide 1MVC_Zinc_XP2.in -HOST xgrid-server etc. 

Finally you take the top-scored 1000 compounds (as we did before for the SP process where we took the 40.000 top-scored ) using the same script (GlideSortScript_ZINC.csh):

/opt/schrodinger2012/utilities/glide_sort -n 1000 -o  file_XP_1000_pv.maegz -r  file_XP_1000_.rept -hbond_cut 0.00 -cvdw_cut 0.00 -metal_cut 10.00 glide-dock_XP_0_pv.maegz glide-dock_XP_1_pv.maegz glide-dock_XP_2_pv.maegz glide-dock_XP_3_pv.maegz glide-dock_XP_4_pv.maegz glide-dock_XP_5_pv.maegz glide-dock_XP_6_pv.maegz glide-dock_XP_7_pv.maegz glide-dock_XP_8_pv.maegz

Some useful tips:

Run faster with the -LOCAL flag:
/opt/schrodinger/glide -LOCAL file.in

path for ZINC on apple:
/Network/Servers/xgrid-Server.xgrid/Volumes/RAID/NetUsers/pgkeka/PI3K/Screening/ZINC/druglike-Zinc-library-mol2/

Apple cluster run-command:
/opt/schrodinger/glide your_XP_file.in -HOST xgrid-server

CSAR is a community resource for Docking and Scoring Development

In their third benchmark competition, over one dozen proteins were designed to bind a steroid. Can you determine which were successful? (timeframe: 1 month and ends on 26 April 2013). See www.CSARdock.org to download the zip file.

This exercise is built around designed proteins developed by Tinberg, Khare, and Baker. This is an interesting twist on traditional docking/scoring problems because there is no existing protein data on which to train models (as was possible in the first two exercises). Three phases, with a possible fourth phase are coming up:

Phase 1: Protein design – Over one dozen proteins were designed to bind a steroid. Can you determine which were successful? (timeframe: 1 month, ending 26 April 2013)

Phase 2: Scoring – Given the set-up crystal structures of the proteins and a set of pregenerated docking decoys, can you identify the correct poses for the steroid? (timeframe: start after Phase 1 ends and run 2 weeks)

Phase 3: Selectivity – Given the set-up crystal structures of some protein-steroid complexes, can you predict the binding affinity (or relative ranking) of several steroids binding to the same protein? (timeframe: start after Phase 2 ends and run 1 month)

Phase 4: Predict the effect of protein mutations – There is extensive mutational data available. Given a subset, can you "bin" their effects into weak, moderate, significant changes in binding? (timeframe: longer and not yet determined)

Riccardo Baron also put together a great computational methods volume you should have access to:http://link.springer.com/book/10.1007/978-1-61779-465-0/page/1
Here's the list of contents:

Drug Binding Site Prediction, Design, and Descriptors
A Molecular Dynamics Ensemble-Based Approach for the Mapping of Druggable Binding Sites

• Anthony Ivetac, J. Andrew McCammon Look Inside Get Access

Analysis of Protein Binding Sites by Computational Solvent Mapping

• David R. Hall, Dima Kozakov, Sandor Vajda

Evolutionary Trace for Prediction and Redesign of Protein Functional Sites

• Angela Wilkins,Serkan Erdin,Rhonald Lua,Olivier Lichtarge

Information Entropic Functions for Molecular Descriptor Profiling

• Anne Mai Wassermann,Britta Nisius,Martin Vogt,Jürgen Bajorath

Virtual Screening of Large Compound Libraries: Including Molecular Flexibility
Expanding the Conformational Selection Paradigm in Protein-Ligand Docking

• Guray Kuzu,Ozlem Keskin,Attila Gursoy,Ruth Nussinov Look Inside Get Access

Flexibility Analysis of Biomacromolecules with Application to Computer-Aided Drug Design

• Simone Fulle,Holger Gohlke

On the Use of Molecular Dynamics Receptor Conformations for Virtual Screening

• Sara E. Nichols,Riccardo Baron,J. Andrew McCammon

Virtual Ligand Screening Against Comparative Protein Structure Models

• Hao Fan,John J. Irwin,Andrej Sali

AMMOS Software: Method and Application

• T. Pencheva,D. Lagorce,I. Pajeva,B. O. Villoutreix,M. A. Miteva

Rosetta Ligand Docking with Flexible XML Protocols

• Gordon Lemmon,Jens Meiler

Normal Mode-Based Approaches in Receptor Ensemble Docking

• Claudio N. Cavasotto

Application of Conformational Clustering in Protein–Ligand Docking

• Giovanni Bottegoni,Walter Rocchia,Andrea Cavalli

How to Benchmark Methods for Structure-Based Virtual Screening of Large Compound Libraries

• Andrew J. Christofferson,Niu Huang

Prediction of Protein-Protein Docking and Interactions
AGGRESCAN: Method, Application, and Perspectives for Drug Design

• Natalia S. de Groot,Virginia Castillo,Ricardo Graña-Montes,Salvador Ventura

ATTRACT and PTOOLS: Open Source Programs for Protein–Protein Docking

• Sebastian Schneider,Adrien Saladin,Sébastien Fiorucci,Chantal Prévost

Prediction of Interacting Protein Residues Using Sequence and Structure Data

• Vedran Franke,Mile Šikić,Kristian Vlahoviček

Rescoring Docking Predictions
MM-GB/SA Rescoring of Docking Poses

• Cristiano R. W. Guimarães

A Case Study of Scoring and Rescoring in Peptide Docking

• Zunnan Huang, Chung F. Wong

The Solvated Interaction Energy Method for Scoring Binding Affinities

• Traian Sulea, Enrico O. Purisima

Linear Interaction Energy: Method and Applications in Drug Design

• Hugo Gutiérrez-de-Terán, Johan Åqvist 

Schrodinger's March 2013 seminar series begin next Tuesday, March 5. This seminar series will include presentations on Phase Shape for ligand-based shape screening and new ensemble docking practices to account for protein flexibility.

Schrodinger is also debuting a Materials Science Suite this year - Dr. Vyacheslav S. Bryantsev of Liox Power will also be presenting on computational modeling of rechargeable Li-air battery materials on March 7. Please feel free to invite your colleagues in the materials industries to join us for this materials-focused presentation!

Here is the full webinar list with dates:

Accounting for protein flexibility in virtual screening using ensemble docking

Dr. Woody Sherman
Schrödinger's Vice President, Applications Science
Abstract March 5, 2013
10 am ET Register
1pm ET Register

Computational Modeling of Rechargeable Li-Air Battery Materials
Dr. Vyacheslav S. Bryantsev
Liox Power, Senior Scientist
Abstract March 7, 2013
10 am ET Register

Using Phase Shape for rapid ligand alignment and virtual screening
Dr. Woody Sherman
Schrödinger's Vice President, Applications Science
Abstract March 12, 2013
10 am ET Register
1pm ET Register

In Maestro when visualizing a structure file, you can  select Tools -> Measurements -> Contacts panel in order to see Good contacts and Bad&Ugly contacts. When contacts are displayed, good contacts are green by default, bad ones are orange, and ugly ones are red. These criteria whether a contact is defined as good, bad or ugly are based on the following formula:

C = D12 / ( R1 + R2 )

where D12 is the distance between atomic centers 1 and 2, and R1 and R2 are the radii of atomic centers 1 and 2. C is defined as the "contact cutoff ratio" in Maestro and has default values of 0.85 for "Bad" contacts and 0.70 for "Ugly" contacts. C must be monotonically increasing for each of the contact types, that is C(ugly) < C(bad) < C(good). These three values then provide 4 ranges. Distances greater than C(good) are not marked, less than C(good) but greater than C(bad) are marked as "good", etc.

This answer was provided by the Schrodinger help team.

Zoe