BrownDye Tutorial

Running on a Linux Computer

This tutorial will walk you through two different types of BD simulations on the Thrombin-Thrombomodulin system, which is an important component of the blood-clotting cascade. Thrombin consists of 295 amino acids, and thrombomodulin consists of 117 amino acids. (My thanks to Adam Van Wynsberghe for the necessary data on these two molecules.)

This tutorial assumes that you have Linux on your computer. If you have only Windows, a quick and easy way to get Ubuntu Linux onto your machine is Wubi (Windows Ubuntu Installer). Other necessary software is the Ocaml compiler, which is free and can be obtained from here . If you are using Ubuntu or a similar Linux based on Debian, you can just type

sudo apt-get install ocaml

You will also need APBS; this can also quickly be installed on Ubuntu by typing

sudo apt-get install apbs

Finally, you can install BrownDye on your machine by going to the website , downloading the file browndye.tar.gz, moving the file to a directory of your choice, and typing

tar xvfz browndye.tar.gz
cd browndye
make all

All of the executables are now in the directory browndye/bin; you need to add this directory to your path. If you use a bash shell, this can be done by typing

PATH=$PATH:fullpath/browndye/bin

where fullpath is the full path name of the directory containing the browndye distribution. If you now want to run the thrombin-thrombomodulin tutorial, go to browndye/thrombin-example.

The atomic coordinates for thrombin and thrombomodulin are in files t.pqr and m.pqr. Because BrownDye works primarily with XML files, you must convert this two files to an equivalent XML format:

pqr2xml < t.pqr > t-atoms.xml
pqr2xml < m.pqr > m-atoms.xml

Next, you must generate the electrostatic grids, in dx format, using APBS:

apbs t.in
apbs m.in

where the input files t.in and m.in are provided. The grids are output to t-PE0.dx and m-PE0.dx. Throughout this session, thrombin will be denoted by the prefix "t" while thrombomodulin will be denoted by prefix "m". Be sure to take note of the Debye length in the APBS output if you don't feel like calculating it by hand; it will be needed later.

In addition to atomic coordinates and grids, the third key input is the set of reaction criteria. This can be generated from the two coordinate files t-atoms.pqrxml and m-atoms.pqrxml, and a file, protein-protein-contacts.xml, which describes which pairs of atom types can define a contact (It is actually stored as protein-protein-contacts.xml.bak; you need to copy from this file.) The program make_rxn_pairs takes these three files and a search distance to generate a file of reaction pairs. This assumes that the coordinates of the two molecules are consistent with the bound state.

make_rxn_pairs -mol0 t-atoms.pqrxml -mol1 m-atoms.pqrxml -ctypes protein-protein-contacts.xml -dist 6.0 > t-m-pairs.xml

The resulting file still is not suitable for input into the simulation programs, however. I've made the program general enough to have more than one reaction in a simulation, so one could envision having several such reaction pair files that would need to combined into a final reaction description file. For now, you can use the program make_rxn_file, which generates an input file for the case of one reaction:

make_rxn_file -pairs t-m-pairs.xml -distance 5.5 -nneeded 3 > t-m-rxns.xml

This generates a reaction description file which tells the simulation programs that if any 3 of the atom pairs approach within 5.5 Angstroms, a reaction occurs.

A note: if you don't feel like typing the above commands in, especially if you make changes and need to do it repeatedly, I have included a Makefile in the example directory. Typing

make all

should run the above commands.

The remaining pieces of information are contained in the file input.xml. (This must be copied from the file input.xml.bak.) It contains, among other things, information on the solvent, information on each molecule, and parameters governing the simulation itself. For the sake of efficiency, the larger molecule should be "molecule0" if there is a large size difference. Also, each molecule is assigned a prefix as mentioned above; these are used in naming the intermediate files that are generated in the next step:

bd_top input.xml

The bd_top program is written in Ocaml using a Unix "make"-like utility that I wrote to help orchestrate the creation of the files. Like "make", if an intermediate file is changed or replaced, running bd_top (which is analogous to a Makefile) will run only those commands necessary to re-generate files that depend on the updated file. Unlike "make", this utility, which I call the "Orchestrator", can also read information from xml files and have chains of dependent calculations (eventually I want to write a version in Python so it will look more familiar to most people). So, when the command is executed, the following files are generated for thrombin:

t-charges.xml - a set of effective charges
t-distances - a dx file containing the distances of each exterior point to the nearest interior point, and vice-versa
t-ellipsoids.xml - bounding ellipoid using for estimating hydrodynamic interactions
t-hydro-radius.xml - hydrodynamic radius
t-q2.xml - effective charges squared; used for the desolvation forces
t-q2-cheby.xml - Chebyshev interpolation data structure for effective charges squared; used for desolvation forces
t-surface.xml - a list of atoms on the surface found by rolling a probe sphere.
t-surface-atoms.xml - actual list of surface atoms to be used for collision detection
t-surface-atoms-hard.xml - list of surface atoms that interact via a hard-sphere interaction
t-volume.xml - list of effective volumes of each atom
t-mpole.xml - multipole expansion of effective charges
ins-t-PE0.xml - grid of points denoting inside (1) or outside (0)

The corresponding files with prefix "m" are also generated for thrombomodulin. In addition, the following file is generated:

m-cheby.xml - Chebyshev interpolation data structure for effective charges; used for Coulombic forces

The following files are generated for both molecules:

t-m-solvent.xml - solvent information
t-m-old-data.xml - previous data, used for detecting changes in parameters
t-m-return-dist.xml - information used for appropriately returning the molecule to the b-sphere; has data on the time and probability distributions.
t-m-simulation.xml - contains additional information generated by the auxilliary programs; is read directly into the simulation programs.

The nice thing about these intermediate files is that any of them can be replaced or changed, and then bd_top run again to update everything. For example, right now I'm using a simple test-charge approximation by default, but one could easily generate effective charges using another program such as SDA, convert the output into the appropriate XML format, replace t-charges.xml and m-charges.xml, and run bd_top again.

Another useability note: if you want to clean things up, you can delete all *.dx and *.xml files; the two xml files that you need to get started are also available as input.xml.bak and protein-protein-contacts.xml.bak.

At this point, you can choose to do a simulation of one trajectory at a time, or you can do a weighted-ensemble simulation. In general, the weighted-ensemble method is not as efficient at the single-trajectory method unless the probability of a reaction event is very low.

To perform the single-trajectory simulation, the following is executed:

nam_simulation t-m-simulation.xml

The results end up in results.xml, as designated in input.xml. As the simulation proceeds, you can look at results.xml to see the simulation progress. (This is called nam_simulation after Northrup, Allison, and McCammon, who came up with the first algorithm of this type. This code uses a fancy variation on the orginal algorithm.).

At any point, you can use compute_rate_constant to analyze results.xml and obtain an estimate of and 95% confidence bounds on the reaction rate constant in units of M/s. The file t-m-solvent.xml must also be given to this program:

compute_rate_constant < results.xml

This will put the rate constant results to standard output.

To perform the weighted-ensemble method, you must first generate the bins for the system copies:

build_bins t-m-simulation.xml

The number of system copies used in the bin-building process are given in input.xml in the n-copies tag. As it runs, reaction coordinate numbers will go scrolling past; they should keep getting smaller and eventually stop. If that does not happen, i.e., the numbers keep on going, you might need to increase the number of system copies, or it might be that your reaction criterion is unattainable. Assuming is converges, the bin information is place in t-m-bins.xml. The actual weighted-ensemble simulation is then run:

we_simulation t-m-simulation.xml

As before, the results are output to results.xml. In each row of output numbers, the right-most number is the flux of system copies that escaped without reaction, while the other ones are reactive fluxes. So, even for a rare reaction event, you should at least see small numbers for the reactive fluxes after the system has reacted steady-state. This can be visually examined at any point, and can also be analyzed as above, but with a different program:

compute_rate_constant_we -sim results.xml -solvent t-m-solvent.xml

Because the streams of numbers are autocorrelated, a more sophisticated approach for computing confidence intervals is used, and if there are not enough data points, the program compute_rate_constant_we will simply refuse to provide an answer.

Ideas for fun

You can change the random number generator seed, under the seed tag. Good to do if you're bored but don't have the energy to do anything else.

One parameter to play with is the reaction criterion distance, which is the -distance input to the program make_rxn_file. The number given in the tutorial and the Makefile is 5.5, but you can change that by re-running make_rxn_file or by changing it in the Makefile.

You can also change the ionic strength in files t.in and m.in and generate new APBS grids. Note: you must take note of the new Debye length and put that value in the file input.xml. So far, it is not possible to automatically get the Debye length from the output DX file of APBS.

If your machine has several processors, you can change the value under the tag n-threads in file input.xml and see it run under several processors. So far, it runs only on shared-memory machines using pthreads. Even on a single-processor machine, you can still run several threads, but it does not make the programs go any faster.

A final useability note: Most of the programs will output a description of themselves and their options if you type in

program -help

Running on an Opal Server

The flow of events is the same as above, except it is done on one of the clusters via a web browser, and for now, the bd_top and nam_simulation steps are combined into one step.

Using the files t.pqr and t.in found in the input files, generate the APBS grid t.dx.
Using the files m.pqr and m.in found in the input files, generate the APBS grid t.dx.
Go to one of the Opal servers at rocce.ucsd.edu or kryptonite.ucsd.edu.
Select the service Browndye: pqr2xml, select the option Input URL, using another browser tab go to the input files, and copy and paste the URL of t.pqr into the Input URL box. Copy and pasting URL's can easily be done by moving the curser over the link, pressing the right mouse button and selecting Copy Link Location.
Click the Submit button
When the page indicates the execution has finished, click on the Output Base URL, inspect the file t-atoms.pqrxml in new browser tab, and do not delete the tab.
Repeat the previous three steps using m.pqr to generate the file m-atoms.pqrxml in its own tab.

The following steps are used to generate the reaction pairs file:

Select the service Browndye: make_rxn_pairs.
Copy the URL of t-atoms.pqrxml from Browndye: pqr2xml service into the box for Molecule 0 Input URL.
Copy the URL of m-atoms.pqrxml from Browndye: pqr2xml service into the box for Molecule 1 Input URL.
Copy the URL of protein-protein-contacts.xml in the input files into the box for Contacts Type URL.
Enter 5.5 into the Search Distance box
Enter t-m-rxn-pairs.xml into the Output File box.
Submit the job, and when finished, inspect the t-m-rxn-pairs.xml file. Do not delete the tab.

The following steps are used to generate the reaction description file:

Select the service Browndye: make_rxn_file.
Copy the URL of the the file t-m-rxn-pairs.xml from the output of make_rxn_pairs to the Pairs Input URL box.
Enter 5.5 into the Reaction Distance box.
Enter 3 into the Number of Required Contacts box.
Enter t-m-rxns.xml into the Output File box.
Submit the job, and inspect the results when finished.



 The following steps are used to run bd_top and nam_simulation

 Select the service bd_top.
As above, copy the URL's for t-atoms.pqrxml and m-atoms.pqrxml into
the boxes for Molecule 0 URL and Molecule 1 URL.
Copy the URL for t-m-rxns.xml into the Reaction File URL box.
Copy the URL's or files for t.dx and m.dx into the APBS boxes
for Molecule 0 and Molecule 1.
Copy the URL of the input file input.xml in the input files into the
Input File URL box.
 Submit the job.
 When the file results.xml finally appears in the Output Base URL,
keep checking it if you want until 1000 trajectories have been run and
the job stops.
 At any time, copy the URL for results.xml into the input URL box
for the Browndye: compute_rate_constant service, and check the file
stdout.txt for the rate constant information.


 References 


Huber, GA and McCammon, JA.  Browndye: A Software Package for Brownian
Dynamics.  Computer Physics
Communications 181, 1896-1905 (2010)


Ermak, DL and McCammon, JA. Brownian Dynamics with Hydrodynamic Interactions,
J. Chem. Phys. 69, 1352-1360 (1978)


Northrup SH, Allison SA and McCammon JA.
Brownian Dynamics Simulation of Diffusion-Influenced Bimolecular Reactions
J. Chem. Phys. 80, 1517-1526


Luty BA, McCammon JA and Zhou HX. Diffusive Reaction-Rates From Brownian Dynamics Simulations - Replacing the Outer Cutoff Surface by an Analytical Treatment,
J. Chem. Phys. 97, 5682-5686 (1992)


Huber GA and Kim S. Weighted-ensemble Brownian dynamics simulations for protein association reactions, Biophys. J 70, 97-110 (1996)


Gabdoulline RR and Wade RC. Effective charges for macromolecules in solvent,
J. Phys. Chem 100, 3868-3878 (1996)