Thursday, March 31, 2011

Discussion of the Mulholland Paper

This is the discussion page for the Mulholland paper titled:

High Level QM/MM Modeling of the Formation of the Tetrahedral Intermediate in the
Acylation of Wild Type and K73A Mutant TEM-1 Class A beta-Lactamase
[1]

General notes on the paper

  • The study is an ongoing project at the Mulholland group. Initial QM/MM calculations was done using AM1/CHARMM22[2] on the very same structure, later gas phase QM corrections were calculated at the B3LYP/6-31+G(d)//AM1/CHARMM22 level of theory[3]. The presented paper, discuss the application of MP2 and SCS-MP2 in the QM region in the field of the MM charges.

  • The structure used in this study is generated by using a mutated (E166N) beta-Lactamase structure (PDB: 1FQG) because of its resolution. The structure preparation was discussed in the 2005 paper by the same group [3]. To obtain the wild type, a single mutation was made based on a lower resolution structure. Hydrogens were added and the whole thing was solvated using a huge sphere (26 Å in radius) of pre-equilibrated water (see discussion below). Once a satisfactory initial structure was obtained, minimization of water, then crystal structure was carried out. Finally, a total of 3284 atoms was obtained (56 QM atoms and 3228 MM atoms) centered around the hydroxylic oxygen on Ser70.

  • As a result, a potential energy barrier is generated based on two reaction coordinates (again, see discussion below). The authors do note, that entropy, thermal and zero-point vibration corrections are for direct comparison to experiment.


Topics of (intense) discussion:


  • Reaction Coordinates: Of the two reaction coordinates used in this study, one of them, Ryz, involved three distances. This could lead to multiple geometries having the same distance parameter. There is no mention in the paper about how this is circumvented albeit the figures looks like everything went fine.
  • To be fixed, or not to be: There is no clear information about how much water and structure is fixed beyond the 18 Å radius during the minimization of the water and/or protein structure. In the 2005 paper, the generation of the structure is explained as "... truncated, by deleting every water molecule and residue, which did not have a heavy atom within 18 Å of the reaction center", where as the later 2009 paper, the statement is "Every atom further than 18 Å away was held fixed". The implications of this, is either a thin shell at a distance 18 Å away or an 8 Å thick water shell around the entire complex.
  • You cannot trust the barrier: As the authors mention themselves, one should add vibrational corrections as well as entropy effects. At least, to confirm the results, another snapshot of the geometry should be made (including all the steps as was done to prepare the structure).


[1] Hermann et al, J. Phys. Chem. A, 2009, 113, 11984
[2] Hermann et al, J. Am. Chem. Soc. 2003, 125, 9590
[3] Hermann et al, J. Am. Chem. Soc. 2005, 127, 4454

Tuesday, March 29, 2011

New COMS member: Thomas Hamelryck

Thomas Hamelryck and his group has joined COMS.  Thomas leads the Structural Bioinformatics Group in the Department of Biology.

Welcome Thomas

Wednesday, March 23, 2011

Gabedit graphical user interface

I found a neat free software called Gabedit. It is a graphical user interface for making/reading input and output files from other softwares such as GAMESS and MOPAC. You can build molecules and send it to any of the programs listed. It can act as a driver for molecular dynamics simulations using MOPAC. If running a MOPAC geometry optimization one can view the geometry and energy of each step, much like gaussview for gaussian. So far I have only managed to get it to work for Windows, so if anyone is able to make the linux version work, please let me know...

Cheers,
Martin

Monday, March 21, 2011

Lovende forskningsmetode åbner kattelem for nye lægemidler


Dansk forsker tæt på at have knækket kompliceret kode for analyse af proteinmolekyler. Det kan i bedste fald betyde, at medicinalindustrien i nær fremtid kan udvikle nye og forbedrede lægemidler.
Forestil dig, at din dørlås til huset ikke vil åbne. Du tilkalder en låsesmed, som ankommer i en lastbil fyldt til bristepunktet med hundredtusindvis af forskellige sæt nøgler.

Hver enkelt skal afprøves, da videnskabelige og tekniske begrænsninger gør det umuligt for låsesmeden at identificere indmaden af dørlåsen og derefter lave en tilsvarende kopi.

Scenariet virker lige så absurd, som det er omstændeligt. Men selve arbejdsmetoden er faktisk almindelig videnskabelig praksis blandt forskere, der beskæftiger sig med de voldsomt komplicerede, men yderst betydningsfulde proteinmolekyleanalyser.

Ved at opnå et faktisk billede af strukturen i proteiner kan forskerne nemlig både hurtigere og bedre hæmme det enzym, der har forbindelse til en given sygdom.

Med andre ord er givtige proteinanalyser forstadiet til at forstå og identificere stort set alle slags sygdomme.

I dag kræver proteinanalyser imidlertid både højt uddannede forskere og enorme mængder tid. Men den langsommelige proces kan nu vise sig at være ovre. En ny strukturanalyse menes at kunne reducere tidsforbruget for et enkelt protein fra ti måneder til to dages computerarbejde.

»Banebrydende resultater«

Det er ph.d.-forskeren Anders Christensen, der står bag den såkaldte kvanteberegningsmetode. Han vurderer, at det reducerede tidsforbrug vil kunne gøre underværker for medicinalindustrien.
»Det vil give medicinal- og biotekvirksomheder mulighed for at undersøge markant flere proteinmutationer. De vil altså have mange flere kandidater at vælge imellem, når de skal udvikle et nyt, virksomt stof,« forklarer Anders Christensen, der netop har taget hul på sin kandidatuddannelse på Københavns Universitet.

Testresultaterne har indtil videre været så lovende, at medicinalgiganten Novo Nordisk er gået ind i projektet, som de støtter med 1,3 mio. kr. over en treårig periode.

Det er ganske klogt, vurderer professor i teoretisk kemi ved Kemisk Institut Jan Halborg Jensen.
»Hvis det lykkes, er det banebrydende resultater for forskningen. Et oplagt eksempel på sygdomme, vi muligvis vil kunne identificere, er Alzheimers og visse cancersygdomme. De er bygget op omkring små proteiner, og hvis vi kan afsløre strukturen på dem, vil det kunne åbne for udviklingen af helt nye lægemidler,« siger Jan Halborg Jensen, der fungerer som faglig vejleder for Anders Christensen.
Ifølge Jan Halborg Jensen opgiver mange forskere allerede at granske proteinopbygningen hos en del sygdomme, da den konventionelle strukturanalyse er enormt kompliceret.

Professoren forklarer, at den nuværende analyseproces svarer til, at man reparerer noget, man ikke kan se. Mange forskere arbejder med andre ord i blinde, når de skal analysere forskellige proteinmolekyler. Men den nye kvanteberegningsmetode giver altså fornyet håb til forskerne, der i øjeblikket kæmper hårdt med at få bugt med de vanskelige og komplekse proteinanalyser.

»Anders Christensens metoder ser spændende ud, og vi hopper da helt sikkert med på vognen, hvis resultaterne også fremadrettet er positive. Proteinanalyser er nemlig ekstremt ressourcekrævende. Det kan i værste fald tage flere år at foretage, mens der er visse enzymer, det slet ikke er muligt at analysere med de nuværende redskaber. Hvis metoden viser sig at holde vand, er det helt klart et kvantespring for forskningen og dermed også et skridt i den rigtige retning mod at udvikle nye lægemidler,« siger funktionschef i forskningen hos Lundbeck Klaus Bæk Simonsen.

Sunday, March 20, 2011

Building molecules by naming them

It turns out you can build-by-name in Avogadro as well.  I have posted a screencast of this feature over at Molecular Modeling Basics.  Please leave any comments there.

Saturday, March 19, 2011

Tuesday, March 15, 2011

How to implement Jmol in Beamer presentations

This week I had to day a presentation for the QM/MM course held by Professor Jan H. Jensen at the University of Copenhagen in which I would like to present the cytochrome P450cam enzyme as an interactive 3D Jmol model. This turned out to be a somewhat tedious project to do in Beamer, mostly due to the fact that I only read the first half of this tutorial section of the Jmol Wiki and hence not the second half which actually addresses many of the issues which I encountered. Note to self: Read all the introductions - the problem is most likely encountered before by someone else and hopefully solved already. Still, I would like to describe how I did it, as I actually ran into some minor TeX problems which I had to solve on my own.

I am running LaTeX thorugh Cygwin DLL 1.7.8-1 on my Windows Vista OS but the points on how to install the additional packages are general and thus apply to LaTeX being run on Mac OS as well as Linux. If you are running LaTeX via MiKTeX or any other program implementation of LaTeX, the packages should be installed without difficulty via the associated integrated package management.

Generate the Jmol input:

  • Note - you need at least version 12.0.RC26 to make .idtf files. Download the latest version here.

I retrieved the cytochrome P450cam (ferrous-dioxygen state) 1DZ8 input from the Protein Data Bank (PDB) here [1]. The .pdb file is downloaded by choosing the 'PDB File (Text)' from the 'Download Files' drop-down list in the upper right corner of the screen.
The .pdb file is opened in Jmol and should now be converted into a Universal 3D file (.u3d) but unfortunately Jmol does not support this as of yet (see the comment section in the previous post on embedding interactive 3D graphics in .pdf files from the MolecularModelingBasics blog here. What Jmol does support, though, is the generation of .idtf files. Once you have your molecular system open i Jmol, open the script console and type in:

write /home/Janus/Documents/test/test.idtf

This generates not only the .idtf file (test.idtf) but also a .tex file called test.idtf.tex which includes the appropriate TeX code.

Conversion of .idtf to .u3d format:

In order to convert to the .u3d format, I used the Universal 3D Sample Software in the newest version (U3D_A_061228_5). After unzipping, I typed in the following within CMD (again, this is Windows but is easily adapted for Mac or Linux OS):

C:\cygwin\home\Janus\presentation_qmmm>"\Users\Janus\Documents\U3D_A_061228_5\Bin\Win32\Release\IDTFConverter.exe" -input test.idtf -output test.u3d

This produces the appropriate test.u3d file and is the file called from the code in test.idtf.tex.

Implementation into Beamer:

I used the Beamer class of LaTeX to create the presentation and quickly realised from the errors in the initial attempts to compile that I needed (at least) two packages in addition to the minimal set of packages that I had already declared [beamerthemesplit, graphicx, amssymb, amsfonts, amsmath, hyperref]. From the test.idtf.tex file I could see that I needed the 'movie15' style which I located and downloaded from CTAN. Once downloaded, I placed it within my local texmf environment and updated:

$ pwd
/usr/share/texmf/tex/latex/misc
$ mkdir movie15
$ mv ./movie15.sty movie15
$ texhash

From the next attempt to compile, I realised from the .log file that I also needed the 'ifdraft' package. I found the ifdraft.dtx file on CTAN here and compiled it again within my texmf environment

$ pwd
/usr/share/texmf/tex/latex/misc
$ mkdir ifdraft
$ mv ./ifdraft.dtx ifdraft
$ tex ifdraft.dtx
$ texhash

Now I was able to compile. I noticed that the test.idft.tex file recommends the use of 'verbatim' but I removed this as I had no need for it. This might not always be the case, though.

Choice of PDF viewer:

I stumbled upon this obstacle when I had finally compiled and attempted to view the .pdf file. I have always been using TeXWorks as it integrates nicely with TeX but I realised that it did not support 3D illustrations and I hence had to download Adobe Acrobat Reader from here. The link points towards the download page for other versions than 9.x and the new X Reader, as none of these showed to be compatible with my 3D illustrations in line with TeXWorks (sigh!). Thus, if you do not have a license for Adobe Acrobat Pro Extended, you will have to use the Reader in a version no newer than 8.2*. I have used 8.2 and it works nicely.

* This might differ for other operating systems. Let me know.

The result:

I have uploaded the result (both the .pdb, .tex, and .pdf files archived as test.tar) on the QM/MM at KU site if you want to look at it or replicate from the .pdb file (here: 3L61 - find it at PDB here [2]).

[1] Schlichting et al. - Science, 2000, 287, 1615
[2] Lee et al. - Biochemistry, 2010, 49, 3412

HowTo create the right environment to use PyRosetta on a Linux box

Since one year I have been involved in a project for which I decided to use a program called PyRosetta. From the website: "PyRosetta is an interactive Python-based interface to the powerful Rosetta molecular modeling suite. It enables users to design their own custom molecular modeling algorithms using Rosetta sampling methods and energy functions." The program comes as pre-compiled binaries for various platforms. The Mac version works fine in an out-of-the-box fashion, most likely because my Mac laptop and the small Mac-Mini where I installed PyRosetta are quite up-to-date. Unfortunately these resources are not sufficient to let me use extensively the program, but I had to struggle a lot to be able to use it on the Linux based clusters we have available here at the Department. Here is how I managed.

The Linux 64bit version of PyRosetta didn't run out-of-the-box, just like that. PyRosetta has a user forum that is only now slowly starting to work, so it wasn't easy to access more detailed information. If you don't have root access to your system, or simply you don't want to mess around with the rest of your installation you will need a local installation of gcc, and Python.

Let's suppose you are installing everything in /home/pyuser/programs . You need first to locally install mpc, mpfr and gmp, following the respective instructions. Download then gcc 4.5.2 or newer and tailor configure it from an object directory with:
../gcc-4.5.2/configure --prefix=/home/pyuser/programs --enable-bootstrap --enable-shared --enable-thread=posix --enable-checking=release --with-system-zlib --enable-__cxa_atexit --disable-libunwind-exceptions --enable-languages=c,c++,fortran --disable-libgcj --with-tune=generic --witharch_32=i586 --build=x86_64-redhat-linux6E --with-gmp=/home/pyuser/programs --with-mpfr=/home/pyuser/programs --with-mpc=/home/pyuser/programs
Then run make and make install. Afterward you will need a local installation of Python 2.6, compiled with the just made gcc. Run the configure command as:
./configure --prefix=/home/pyuser/programs --enable-shared LDFLAGS='"Wl,-rpath /home/pyuser/programs/lib" --enable-unicode=ucs4
The directory /home/pyuser/programs/lib must be already present prior to the configure command, even if empty.

Then run make and make install. Once done, add /home/pyuser/programs, /home/pyuser/programs/bin to $PATH and /home/pyuser/programs/lib, /home/pyuser/programs/lib64 to $LD_LIBRARY_PATH before any other previously present path. You can check if your installation is correct by typing python. The output should look like something similar to this:
Python 2.6.5 (r265:79063, Jan 24 2011, 16:57:13)
[GCC 4.5.2] on linux2
If everything worked fine then you are good to go, just remember to source SetPyRosettaEnvironment.sh.

One final note: these instructions are relative to PyRosetta 1.0. Recently a 2.0 version has been announced as a Beta, based on the recently released version 3.2 of Rosetta. Unfortunately once again the only indications about which are the system requirements consist in "standard linux/Mac distributions".

Discussion of the Thiel paper

The discussion page for the Thiel paper:

'Coupling and uncoupling mechanisms in the methoxythreonine mutant of cytochrome P450cam: A quantum mechanical/molecular mechanical study.'

Collection of the various points mentioned. Suggestions, feedback, corrections, additions and comments are most welcome.

General notes on the paper:

  • The present mutagenesis study wishes to investigate the effect of the Thr252MeO-Thr mutation in terms of disruptions of the proton relay channels (Asp251 and Glu366) which are essential prerequisites for the conversion of Cpd0 into Cpd1 in the cytochrome P450cam enzymatic cycle. The article thus continues the line of computational mutagenesis research studies initiated by the same group in 2009 [1].
  • The study performs a single classical MD simulation from an initial PDB structure (1DZ8) obtained from group of Schlichting et al. [2] and a single (random) snapshot is taken as initial structure for the QM/MM optimizations. Hydrogen link atoms and the charge shift model were employed to treat the QM/MM boundary and an electronic embedding scheme [3],[4] was adopted in the QM/MM calculations. In the QM calculations, interactions with MM charges were incorporated into the LCAO-MO one-electron Hamiltonian, thus allowing QM polarization, and the QM/MM electrostatic interactions were evaluated from the QM electrostatic potential (at HF/6-31G(d) level) and the MM partial charges.
    The procedure is hence sequential; the QM region is initially optimized with the surroundings fixed (1 QM step) followed by a MM response to the QM region which is now being held fixed. This is again followed by a new QM step (energy and gradient evaluation for the core) and the iteration cycle continues until convergence [5]. Transitions states may thus be located from the inversion of the QM explicit Hessian by the P-RFO module of the HDCL procedure within ChemShell. P-RFO updates the Hessian matrix via the quasi-Newton BFGS optimization algorithm. The QM calculations were done with UB3LYP in Turbomole as a compromise between the need for incorporation of static electronic correlations effects and the large size of the QM region while all the MM calculations were done with the CHARMM22 force field either in CHARMM (MD) or in DL-POLY.
  • The results of the study show that the computed QM/MM barriers indicate unfavorable uncoupling in the case of the Thr252MeO-Thr mutant (unlike when X = Val, Ala, Gly [1]), whereas there are two energetically feasible proton transfer pathways for coupling; these are Mechanism I: homolytic O-O bond cleavage followed by coupled proton-electron transfer and Mechanism II: proton-assisted heterolytic O-O bond cleavage. The study shows Mechanism I to be favorable.
    The corresponding rate-limiting barriers for the formation of Cpd1 are higher in the mutant than in the wild-type enzyme. These findings are consistent with the experimental observations that the Thr252MeO-Thr mutant forms the alcohol product exclusively (via Cpd1), but at lower reaction rates compared with the wild-type enzyme.
    With respect to the two different channels, the rate-limiting barriers are somewhat lower in the Glu366 channel than in the Asp251 channel. The Asp251 channel is in contact with bulk water, though, so it should be rather facile to reprotonate Asp251 after each coupling reaction that involves proton transfer in the Asp251 channel. This is not true for Glu366, which resides in a hydrophobic pocket and is thus difficult to reprotonate.

Topics of discussion:

  • We discussed the potential problems associated with only employing a limited number of MD simulation snapshots in the subsequent QM/MM calculations. This procedure surely does not incorporate effects affiliated with the dynamics of the system, i.e. effects arising from conformational fluctuations are neglected as the span of the configurational phase space is largely limited. The Thiel group includes a short assessment of this issue as they address another (randomly chosen) snapshot in the same manner and achieve similar results, thus confirming the validity of the procedure.
  • At the lecture we discussed this aproval of validity of the computational protocol - which is also build upon the excellent results with respect to experimental references which the group has obtained in previous studies [6],[7] - and we all felt that the neglect of dynamics in terms of the limited amount of snapshot used may potentially cause problems but we all acknowledge the apparent accuracy of the method. We spoke further on the neglect of solvent effects as the hydrogen bonding pattern in the solvent might change dramatically within the time period of the MD simulation potentially leading to increments in energy barriers. In defence of the present study, the active site of cytochrome P450cam resides within a hydrophobic pocket of the enzyme but fluctuations (+/-) in the energy barriers as a result of the shifting hydrogen bonding patterns in the solvent may still occur.
  • Finally, we questioned wether the Pople style basis set, 6-31G, i.e. without polarizations functions, would be efficient in describing the electronic environment within the QM calculations.

Selected references:


[1] Altarsha et al. - J. Am. Chem. Soc., 2009, 131, 4755
[2] Schlichting et al. - Science, 2000, 287, 1615
[3] Altun et al. - J. Phys. Chem. B, 2005, 109, 1268
[4] Bakowies et al. - J. Phys. Chem., 1996, 100, 10580
[5] Turner et al. - Phys. Chem. Chem. Phys., 1999, 1, 1323
[6] Schöneboom et al. - J. Am. Chem. Soc., 2002, 124, 8142
[7] Schöneboom et al. - J. Phys. Chem. B, 2004, 108, 7468

For further reading, please see the list of references at the end of the slide show.

Sunday, March 13, 2011

Building molecules on the web

You can now build molecules on the web simply by typing their names.  For more information see my post on Molecular Modeling Basics.  If you have comments please leave them over at MMB.

Friday, March 11, 2011

Screencasts for everyone


The Education at Its Best initiative has generously provided us with a new Mac.  I chose a Mac because of ScreenFlow, which only runs on Macs.  ScreenFlow is perhaps the best software package for making screencasts, i.e. movies that record what happens on the screen.  I have found screencasts very valuable tools for education in general and for software-related instructions in particular.

Here are two screencasts I made.  One is very simple and shows how to use this website


The other screencast is a little more advanced, but gives you a hint the kind of editing that is possible with the program.



If you are interested in learning how to use ScreenFlow let me know, I'll be happy to show you.  (Of course, ScreenFlow comes with excellent how-to videos).

If you want a login account on the Mac please ask Casper.

Thursday, March 10, 2011

The scientific tax return

Every year I, and other permanent research staff at the Faculty of Sciences, must fill out a "scientific tax return" (the Faculty's name for that, no mine), which is an online form with a bunch of questions.

I have included the questions related to publications below.   Without going into whether such a survey is a good idea at all, there are some messages that this survey sends (assuming one worries about this at all):

1. Don't publish in new journals (they have an impact factor of 0).
2. Don't publish in journals that are not indexed on the Web of Science (they don't contribute to your citations and have an impact factor of 0).
3. Blog posts don't count (I have written 7 blog posts that each were viewed more than 1000 times in the last 12 months)
4. Software doesn't count (for example the PROPKA webserver has been accessed 20,609 in the last 12 months)

Now that we have this thing called the internet, scientific communication and impact, has become so much more than conventional journal articles.

On a positive note, I do like that they ask many different questions, because that gives a better picture about productivity than a single number.

Questions Related to Publications
* Total number of peer reviewed publications published in 2010 (please do not include publications that were still "in press" by the end of 2010)

* Total number of peer reviewed publications published in 2010 included in Web of Science. See manual below.

* Total number of monographs published in 2010 (please do not include publications that were still "in press" by the end of 2010)

* Sum of Journal Impact Factors of all publications published in 2010, see for this table for present JIFs and WikiPedia for more information.

* Number of citations in 2006-2011 for publications from the period 2006-2010, including both years. See manual below.

* Total number of citations for your entire academic career. See manual below.

* Total number of publications from 2006-2010, including both years.

* Hirsch-Index. Hirsch-Index for your entire professional career (XXXX-2011). See WikiPedia and instructions below.

The form had a comments section.  Here is what I wrote.
Publication/Scientific Communication:
* I published the book Molecular Modeling Basics, which has sold 461 copies in 2010.
* I continued to run the blog molecularmodelingbasics.blogspot.com, which received 37,700 visits in the last 12 months.

Wednesday, March 9, 2011

Journalist: force scientists to blog

A journalist from Berlingske (a major Danish newspaper) has, on his own blog (in Danish, sorry!), suggested that researchers should be forced to blog.  Be sure to read the comments.

Here's one response from another blogger (also in Danish, sorry!). 

Tuesday, March 8, 2011

Open PhD Position in Theoretical Biochemistry at the University of Nancy, France

A PhD position (three years) is available starting September 2011 in the Theoretical Chemistry and Biochemistry group at the University of Nancy, France. The successful candidate will work on the development of novel approaches in quantum chemistry and molecular dynamics to describe the vibrational properties of biological systems in interaction with the environment.

The position is funded by the French Ministry and it might include teaching. The position provides access to the French Social Security and Health System.

The successful candidate will have a Master in Theoretical Chemistry or Physics, and a robust computational background. Applicants must show ability to efficiently work in a team and good communication skills.

Research Project

We are presently working on the development of a molecular dynamics approach based on an electronic Hamiltonian at the quantum semi-empirical level. This method can be coupled with linear-scaling algorithms to improve the performance of the calculation. Our approach has been successfully tested on simulations of liquid water and on a model peptide in aqueous solution. Further extensions of the method need to be designed and implemented in order to treat the lateral chains of peptides, and to finally model larger systems, such as proteins. Our goal is to apply this methodology to the study of IR spectra and to vibrational energy relaxation for biological systems. Charge (proton) transfer, as well as the mutual solute-solvent polarization, have to be taken into account to correctly model these phenomena. Different strategies (classical molecular dynamics, Car-Parrinello, QM/MM)will be compared to our original method. Part of the project will be carried out in collaboration with the Chemistry Department at Murcia University (Spain) and with an experimental group (LCPME, Nancy).

Living in Nancy

Nancy is a pleasant city in Lorraine, central France, within a hour and a half by train from Paris, Strasbourg, Germany and Luxembourg. The city center hosts a few UNESCO's World Heritage sites, and it is world-wide known for buildings and masterpieces in the Art Nouveau style. Easy access to hiking, biking, skiing, climbing and other outdoor sports is a great advantage of leaving in Nancy, a city surrounded by woods and by the countryside and only 80 Km away from the Vosges mountains (
http://www.tourism-lorraine.com).

Contacts

Francesca.Ingrosso@cbt.uhp-nancy.fr
Gerald.Monard@cbt.uhp-nancy.fr

Saturday, March 5, 2011

My first attempt at e-publishing

I have made an electronic color supplement to my book.  Read more on Molecular Modeling Basics

Friday, March 4, 2011

NanoGeoScience competences and programs added

The NanoGeoScience group has now added some new programs under "resources", which can be found at https://sites.google.com/site/ccomsatuc/resources.

We believe that of particular interest to others at the Chemistry department at KU is:
Efficient predictions of free energies of solvation, pKa values, solubilities, etc with density functional theory in implicit solvent. Accurate parametrizations for pKa predictions (<0.5 pH units) exist for acids in water, DMSO, acetonitrile and heptane as well as for bases in water, acetonitrile and THF. Equilibrium constants for reactions in solution are also easily calculated.

Other things we do include:
Plane wave density functional theory - for calculations involving crystalline materials in the bulk or at surfaces, or other densely packed systems
Linear scaling density functional theory for very large systems (up to 10000 atoms)
Molecular dynamics of e.g. mineral/liquid interfaces

Cheers,
Martin

Potential Summer Scholarships with Noel O'Boyle

An interesting opening for summer holiday coding with Noel O'Boyle has been published: http://www.ichec.ie/education_training/summer_scholarships/. An excerpt from the homepage regarding parallelization of OpenBabel:

OpenBabel: optimisation of the MMFF94 routines using OpenMP and OpenCL

At the heart of many computational chemistry modeling suites are the forcefields and their associated energy and gradient functions. One such forcefield, MMFF94 (Merck Molecular Force Field 94), was developed by Thomas A. Halgren at Merck and has become one of the most widely used in the drug design domain.

OpenBabel is an open source cheminformatics project and MMFF94 is one of many forcefields it provides. OpenBabel has been downloaded over 164,000 times and is used by over 40 related projects.

This project will involve the optimisation and parallelisation of the MMFF94 functions within OpenBabel and will be conducted in collaboration with Dr. Noel O'Boyle at University College Cork.

Definitely interesting if one has no other business in the summer holidays!

Tuesday, March 1, 2011

Discussion of the Friesner Paper

The discussion page for the Friesner Paper:

QM/MM of P450 BM3

Collection of the various points mentioned. Suggestions, feedback, corrections, additions and comments are most welcome.

General Statements of Interest:
  • The study does not carry out any MD simulation in order to obtain a minimum energy path or a potential of mean force. Instead what is presented is an adiabatic scan of the reaction coordinate. What does this imply: in the transition state search and in the optimization of the reactants/products, the system is optimized into the next local minimum. So it is implicitly assumed that the difference in energy is caused only by the difference in energy of the active site part of the enzyme. Still it is possible that the transition state (TS) system is in a different local minimum than the reactant and therefore calculating the energy barrier from these two system is somewhat questionable. In the present case, the reaction is characterized by a simple hydrogen transfer, so maybe it is a safe assumption that the two systems indeed do not differ for parts other than the active site.
  • TS search: QSite provides the user with three ways of locating TSs: something similar to CalcFC, something like QST2/3 and a general Newton-Raphson minimization (most likely with some sort of Hessian approximation). It is not clear, which method the authors used to compute the TS.
  • This point was brought up by Janus. The "message" of the paper is that the induced fit docking (IFD) procedure provides a reasonable starting structure for the reactant species, as confirmed by the REMD. The paper does not claim to report results which aim at reproducing experimental results for the rate constant.
Considering the Boundaries:
  • The method is defined such that part of the solvent molecules (the outer 4Å of solvent) are being kept fixed during the optimization. The question was, if this is fair to do? As I can tell from my own work, it can happen that in going from one point on the reaction coordinate to another, the system finds a new, different local minimum and a significant (unrealistic in the sense of "this would not happen if the modeling was more sophisticated") change in energy is observed. This is somehow related to the first point above.

Open Points / Questions:
  • We were briefly discussing the way the optimization is being carried out. Apparently, the MM system is being minimized for every QM step of optimization. So no only when the QM gradient is below the threshold. Was there something more to this other than it just being stated?

Jan's presentation for Research Day


Tomorrow the Department of Chemistry is arranging a "Research Day" - a sorta get to know each other thing, aimed at starting new collaborations.  Each participants gets 7 minutes.  Here are my slides.  If nothing else, it is a good chance to advertise COMS and our Education at Its Best project.