CHARMM Development Project Forums User Discussion & Questions Parameter Set Discussion ParamChem: Automation of CGenFF

Hello,

We are happy to announce the availability of the ParamChem website at https://www.paramchem.org/ . The current implementation of ParamChem features the CHARMM General Force Field (CGenFF) program for automatic atom typing and assignment of parameters and charges by analogy. The utility is based on CGenFF, the extension of the CHARMM additive force field that allows for the treatment of drug-like compounds, and allows users to upload molecules in MOL2 format from which a CHARMM compatible RTF of the molecule, including charges, is created along with parameters that supplement CGenFF as required for the molecule. Note that penalty scores are assigned to each charge and parameter produced by the CGenFF program, allowing users to gauge the quality of the generated parameters. However, given the limited transferability of empirical force fields, even parameters with small penality scores may not be optimal and validation of the parameters, as described in the CGenFF manuscript, is highly suggested. Efforts to develop more automated methods for parameter validation are ongoing as part of the ParamChem effort.

We hope you find the ParamChem/CGenFF combination of utility. Details of how to use ParamChem are below.

Alex, Kenno and the entire ParamChem team ( https://www.paramchem.org/AboutUs.html )

• 1) Register at https://www.paramchem.org/Registration/index.php . You will receive an e-mail with information on how to activate your account. Presently, only academic and government users are allowed.
• 2) Upload your molecule of interest in mol2 format at https://www.paramchem.org/AtomTyping/index.php . It is important that all hydrogens are present, in the correct protonation and tautomeric states, and that the connectivity is correct. The xyz coordinates are not critical. The maximum number of atoms is 256, and there is a limit of 100 molecules per user per week. Note that CGenFF should not be used for biological macromolecules; see the notes on the main CGenFF page or the FAQ for details. The highly optimized CHARMM force field for Proteins, Nucleic Acids, Carbohydrates and Lipids can be downloaded freely from the CHARMM force field page and can readily be combined with CGenFF.
• 3) If the interface returns warnings, look at them carefully - warnings often imply that the parameters may not be sufficiently accurate for use without validation and/or optimization.
• 4) Download the str file and open it in a text viewer or editor. In particular, look at the "penalty scores" associated with the partial charges and parameters. Penalties between 10 and 50 indicate that some basic validation is recommended; penalties higher than 50 usually are usually associated with parameters or charges that need additional optimization. It should be noted that even in cases were the penalty score is low, the accuracy of the parameters cannot be 100% guaranteed. In all cases it is recommended that the user perform some level of validation of the parameters prior to use.
• 5) The stream file produced by the CGenFF program represents an extension of the main CGenFF topology and parameter files, which can be downloaded here. Verify that your stream file is compatible with the CGenFF version you downloaded (all the files contain comments with version information). Using incompatible files may not result in errors but may give rise to inconsistent energetics! At the time of writing, the available version of the CGenFF program is 0.9.0, which requires the latest version of the CGenFF force field, 2b6.
• 6) Congratulations, you're ready to run CHARMM. Here's a basic example showing how to run an energy minimization.

Note: Both CGenFF and the functionality exposed through this interface ("the CGenFF program") are in beta stage of development. Although extensive testing has been performed during the alpha stage, we are not responsible for potential problems although we will certainly attempt to address issues associated with problems via email to help@paramchem.org. The CGenFF force field and program will be updated in tandem at regular dates. These updates will increase the number of supported chemical groups, but may also lead to small changes in the charges and parameters assigned to a given molecule. Therefore, when citing CGenFF/Paramchem, it is important to include the version number of the interface (0.9.0 at the time of writing) and the force field (2b6 at the time of writing).

Cite CGenFF/ParamChem respectively as

• CGenFF: K. Vanommeslaeghe, E. Hatcher, C. Acharya, S. Kundu, S. Zhong, J. Shim, E. Darian, O. Guvench, P. Lopes, I. Vorobyov, A. D. MacKerell Jr., J. Comput. Chem. 2010, 31, 671-690.
• Atom typing and assignment of parameters and charges by analogy: K. Vanommeslaeghe et al, in preparation.
• ParamChem interface: www.paramchem.org

We applied a minor update to the CGenFF program - the new version is 0.9.1 . This version has improved functionality for renumbering atoms with duplicate names; the renumbering algorithm in version 0.9.0 in rare cases behaved incorrectly. It also has some minor improvements in the warning and error messages and in the treatment of spiro compounds. Just like version 0.9.0, version 0.9.1 of the CGenFF program requires version 2b6 of the CGenFF Force Field.

I have just tried out ParamChem. I would like to have parameters for a fluorogen molecule which contains a ring with a nitrogen and a sulfur atom. The PDB has no hydrogen atoms. I get the warning:

sp-hybridized aromatic atom

The FAQ says that this warning is usually due to the PDB not containing hydrogen atoms. Any suggestion how I could add hydrogens to my PDB before optaining the PSF?

Thanks!

Please start a new thread for a new question rather than replying to an old one!

Edit: Also, please google (and/or use the forum's search function) before asking a question. Your question has been discussed on this forum just a day before you posted it.

Oops, sorry, I should have used "Search"

Thanks for pointing me to the right answer.

We applied a major update to the CGenFF program - the new beta version is 0.9.6 . The output of this new version should only be used with the new 2b7 release of the CHARMM General Force Field. Version 0.9.6 of the CGenFF program features the following improvements:

• Improved parameters and penalties for substituted planar rings and conjugated double bonds
• "Guess bond orders from connectivity" feature
• Improved atom typing of polyaromatic heterocycles
• Improved accuracy for ketals, neutral enamines and beta-substituted enals

It is strongly recommended to read the detailed description of these improvements below (also available on the changelog page). If you cannot do so right now, at least note the following caveats:

• Using the new "Guess bond orders from connectivity" feature is risky and should be avoided whenever possible
• Resubmitting existing molecules to the new version will often give rise to higher penalties. This does not imply that the new parameters are less accurate; rather, it means that the old penalties were too low.

Detailed description of the improvements in 0.9.6
More accurate parameters and penalties for substituted planar rings and conjugated double bonds
After months of beta testing, it was discovered that in some cases, molecules with substituted planar rings or conjugated double bonds yielded unreasonable parameters with misleadingly low penalties. This turned out to be a fundamental shortcoming in the parameter assignment algorithm, necessitating the implementation of substantial new functionality, which is now available in version 0.9.6. The problem was especially pronounced for angle parameters in substituted planar 3- and 5-membered rings. If you have encountered incorrect behavior in such molecules, or molecules with conjugated double bonds, it is advisable to submit the affected compounds to our new version, and contact us if the problem persists or if anything else is odd. Remember that our software is still in beta stage, and its quality depends on your feedback!
That said, note that the new version will often give you different parameters with substantially higher penalties. This is normal and does not imply that the old parameters were more accurate; on the contrary, it means that the penalties for these parameters were too low in the old version, and that after adjusting them upwards, better (ie. lower-penalty) parameters turned out to be available.

"Guess bond orders from connectivity" feature
It has been brought to our attention that some popular free molecular editors (most noteworthy VMD) produce mol2 files with invalid bond orders. Also, distorted pdb files are frequently submitted, causing OpenBabel to guess the bond orders incorrectly. To work around these problems, the new version of the interface features a check box to instruct the program to discard input bond orders and guess the bond orders from the connectivity. We would like to stress that this feature should be used with EXTREME CAUTION! Ordinarily, the program checks whether the input bond orders are consistent with the connectivity, and returns an error if this is not the case. Doing so has already caught many problems that would otherwise lead to incorrect parameters. However, this check is disabled by choosing "guess bond orders..." This implies that if, for instance, hydrogen atoms are missing or placed incorrectly, the program may output grossly inaccurate parameters without warning. Additionally, guessing bond orders may not always work on 3rd-row and heavier elements.

Improved atom typing of polyaromatic heterocycles
In some cases, polyaromatic heterocycles containing positive nitrogen atoms and/or carbonyl groups received incorrect atom types. This bug is fixed in version 0.9.6.

Improved accuracy for ketals, neutral enamines and beta-substituted enals
This is the result of our ongoing extension of the CHARMM General Force field.

Finally the papers associated with the CGenFF program for atom typing and assignment of parameters and charges by analogy are available online:
K. Vanommeslaeghe, A. D. MacKerell Jr., J. Chem. Inf. Model., DOI: 10.1021/ci300363c
K. Vanommeslaeghe, E. P. Raman, A. D. MacKerell Jr., J. Chem. Inf. Model., DOI: 10.1021/ci3003649

We released another update - the new beta version is 0.9.7 . The output of this new version should only be used with the new 2b8 release of the CHARMM General Force Field. Version 0.9.7 of the CGenFF program is mainly a "monster bugfix release", in which we corrected almost all of the user-reported and self-discovered issues that accumulated over the last 2 years of beta testing. Additionally, we added support for a few important functional groups. Here is a full list of improvements:

• New support for saturated 4-membered rings
• Improved support for epoxides, aliphatic nitro groups, neutral organic sulfonates, sulfamates, benzoic acid esters and derivatives
• Better charge (and parameter) assignment on positive amines
• Residue IDs (RESI) now truncated at 6 characters for improved compatibility
• Fixed penalty score for ranking resonance structures
• Better guessing of bond orders of sulphur and phosphorus-containing moieties
• Better aromaticity perception on some sulfur-containing rings
• More sophisticated aromaticity perception algorithm
• Improved parameter assignment on systems with more than 2 conjugated aliphatic rings
• Correct atom typing when given incorrect bond orders on nitro groups, negative carmabates
• Corrected atom typing on sulfoximines, phosphorimidates, sulfenates
• Fixed formal charge on positively charged bridging nitrogens conjugated to another nitrogen
• Maximum number of atoms increased to 384
• Fixed atom renumbering
• "Penalty" is now uppercase when including parameters that are already in CGenFF
• Clearer warning messages about sulfinamide and sulfurous diamide not being explicitly supported
• Tweaked warning and error messages for nitrates
• Miscellaneous tweaks in assignment of charges and parameters by analogy

It is recommended that users read the detailed description of these improvements in the changelog. In particular, if you are currently using paramchem-generated molecules in an active project, it might be a good idea to check whether your molecule gets the same atom types and total charge in the new beta version. That said, there is no reason to panic; we estimate that less than 1 in 1000 molecules are seriously affected by the bugs that are now fixed.

This is an announcement thread. As mentioned just a few posts up, in the Parameter Set Discussion guidelines and in the general CHARMM forum guidelines, Please start a new thread for a new question rather than replying to an old one! I feel like a broken record here; what would it take for people to start following the rules?!

Also, what you're asking is not a parameter set question, but a technical problem with the hosting server. The group that is responsible for this does not follow these forums! So, while an issue like this is obviously very important to me, and I appreciate your feedback, if you use the appropriate support channel next time, your problem will likely be resolved far quicker.

Edit: technical problem is fixed now, and as a bonus, we used the opportunity to raise the weekly limit to 150. No need to reply to this; let's keep this thread clean.

Edit 2: since a forum admin seems to have (re)moved the post I replied to (thank you for that), we probably want to do the same thing to this post. I feel neither the increased limit nor the transient issue and its resolution merit a formal announcement.

We released another update: 1.0.0 , the first non-beta release. The output of this new version should only be used with the new 3.0.1 release of the CHARMM General Force Field. The reasons we now consider the CGenFF program and force field officially out of beta are:

• very few bugs popped up during the last year;
• we now support a sufficiently large majority of drug-like molecules for routine use in Computer-Aided Drug Design projects.

This also implies that we're open to providing machine-friendly access to nonprofit collaborators; please e-mail us (i.e. do not use the forum, AND ESPECIALLY NOT THIS THREAD) if you are interested. For-profit users may obtain the CGenFF program from SilcsBio, LLC.

Version 1.0.0 of the CGenFF program improves upon 0.9.7.1 by introducing support for a number of important functional groups and improving support for a number of moieties. Here is a full list of improvements:

• Introduced support for "bipyrrole-type" sp²-sp² single bonds between 5-membered rings
• Fixed bug where in rare cases, atoms were incorrectly recognized as "bipyrrole-type"
• Introduced support for "bipyridine-type" nitrogen-containing single bonds between 6-membered aromatic rings
• Greatly improved nonbonded representation of terminal alkynes
• New support for bicarbonates/hemicarbonates, sulfamate anions, sulfamides and isocyanates
• Improved support for semi-flexible 5-membered rings and gamma-lactones, non-amide carbonyls in 6-membered aromatic rings, heteroaromatic bicycles and biphenyls, thioether substitutions on heterocycles, taurine amides, N-benzylamides, 1-alkyl-1,2,3-triazoles and phosphoramidates
• Improved support for neutral sulfamates
• Substantial improvements in charges on aromatic heterocycles
• Reduced numerical noise in charge assignment

It is recommended that users read the detailed description of these improvements in the Change Log.

Finally, we would like to use this opportunity to draw your attention to the fact that We recently published a paper on our "lsfitpar" program for robust fitting of bonded parameters. The program itself will be available under an open-source license in the week of the 26th of April on the MacKerell lab's force field development page. While this is not a GUI-driven interface that automatically takes care of all steps of the parametrization process, we expect it to be very helpful to moderately experienced computational chemists, and to become an important part of the sprawling ecosystem of automatic parametrization interfaces.