Re: Force field design

From: R.J. Swett (
Date: Wed Jul 23 2014 - 23:03:48 CDT

I am a rare contributor to these online discussions, but I'd like to say
that Axel has a point about scientific curiosity and independent
learning. Abhishek, there is very little you cannot learn through
exploration of the manuals, code, and direct experimentation. That is
one of the joys of computational chemistry, if you fail, you have still
learned something (and likely did not burn your eyebrows off in the
process!). Try again. Fail again. Read the errors. Figure it out. Rather
than resorting to asking directly for help and training, I urge you to
1) read thoroughly all the fundamental literature on the scientific
process you wish to pursue, be it parameterization, MD, QM, etc. 2) read
THOROUGHLY the documentation on the software. It is exceptionally rare
that a use case will come up that could not have been answered by
actually READING the manual before beginning. 3) Google your errors.
Don't ask here automatically. If your run fails, google it. If your job
gives junk results, google it. If you don't know how to answer your
question, google it. Between the mailing lists and google scholar, there
is very little that you won't find readily available. Particularly as a
beginner in the field. Remember, this is science, EXPERIMENT! The great
thing about computational chemistry is that you're not burning through
consumables. You are limited by your available computational resources
and your own time only. Making horrible mistakes are an excellent way to
really learn how NOT to make mistakes.

Basic advice: Before you begin ANY experiment, find several papers that
apply some version of that methodology to answer a scientifically
relevant question and read their methods section. Then read all the
references in those methods sections. I am not exaggerating. The
fundamental literature and methods are key. Be sure you understand them.
If you don't, find a review article on the topic and read it. And then
read the references in the sections you don't understand. If the methods
section works through math you don't quite get, take pen to paper and
see if you can work through either the algorithm or implementation
scheme. Performing experiments you don't understand is just as dangerous
here as in an energetic materials lab.

It seems as though you do not have a computational advisor. Whatever
facility you are associated with will likely have someone with
experience in computational chemistry and I urge you to seek them out.
While the above process will educate you, there are innumerable minor
pitfalls that you may be susceptible to. From your email, it's apparent
that you are in the pharmacy school of the university of maryland. This
school has it's own biophysics division. and I count at
least six faculty performing molecular dynamics just from a cursory
glance. I strongly suggest that you speak to your current PI about
training with some of these faculty members who would be able to lend
their expertise if they were so willing. It is inappropriate and
downright rude to demand training and insult the folks who build,
support and develop NAMD when it's clear that you are not using either
your own elbow grease or the locally available resources to learn.

"I am agree few of the points but not all, their are many users like me
using NAMD and try to learn it by online resources available, and I
think most of them face these problems during their work that I am
facing right now."

In response to this, yes, there are many users trying to learn NAMD
through the online resources and yes, it's hard and takes motivation. If
you truly want to learn how to use MD as the powerful tool that it is,
consider the mailing list an absolute last resort. Molecular Dynamics is
it's own field, not something that can be picked up over a weekend and
certainly not something that is to be lightly approached. Before you ask
questions of the people who have spent multiple DECADES of their life
working on this very complex field of study, be absolutely certain that
this is not a question that arises due to misuse or ignorance.

R.J. Swett
Wayne State University
357 Chemistry
Detroit, MI 48201

Lab Phone 313-577-0552
Cell Phone 906-235-0768

On 7/23/2014 11:06 PM, Axel Kohlmeyer wrote:
> i would like to add to this that i am training people in some form of
> computational science for over 20 years now and if i have learned
> anything in the process, it is that most people are not ready to train
> themselves. it is an absolute necessity to have people "get their
> hands dirty" under supervision and have them do exercises where they
> make mistakes and then show them how to figure this out through
> discussions.
> i cannot recount how many times i have been asked to write a book to
> collect all what i have taught people, yet my answer is that it would
> be a mostly useless. even 20 years ago, when things in computational
> science were very different (slower, less complex, less pressure and
> more time to explore and learn things properly) only a limited number
> of people were able to teach themselves without feedback from
> experienced people. i got lucky and had a fantastic supervisor and got
> trained "the hard way".
> despite people being much more exposed to information technology these
> days, computer literacy (in the old-fashioned way that i got trained
> in) is at a low point. our education system is focused on teaching
> people content and not problem solving skills and common sense. as a
> consequence, i have to "dumb down" the content of the courses,
> tutorials, or schools that i am teaching at in order to motivate
> people to actually try stuff in a way that they have to reflect, think
> and understand the solution to a problem and not only consume what is
> talked about and blindly repeat what they are told to do. the latter
> point is where the demand for better online teaching material is
> missing the point.
> there is a sufficient amount of material available already.
> axel.
> On Wed, Jul 23, 2014 at 7:20 PM, Kenno Vanommeslaeghe
> <> wrote:
>> I think Axel is right; you really need to talk with your PI about spending a
>> few months in an MD lab. Universities and granting agencies commonly have
>> funds for a grad student to do so, so he has everything to win and nothing
>> to lose. And that goes double for you.
>> As for your complaint about perceived lack of online resources, to continue
>> with the analogy from my posts on the CHARMM forums (do read them), if you
>> buy an NMR or X-ray crystallography machine, the vendor will provide you
>> with manuals and training to allow you to use the machine at a *technical*
>> level. However, this alone will *not* allow you to use the machine to solve
>> scientific problems. For that, you also need:
>> - scientific background
>> - the know-how to combine you scientific background and technical skills
>> into effective research and valid/relevant results
>> Scientific background you can get from university courses and textbooks, but
>> there's no way to obtain know-how but to learn it form a more senior
>> researcher. No vendor in the world provides training for *that*. Also, you
>> don't seem to realize we are not vendors; we are academics. Our main
>> priority is to publish, and even though you may not experience it as such,
>> we're only helping the community out of kindness. But, we cannot help the
>> "many users like you", especially not over e-mail; that's like trying to
>> empty an ocean with a teaspoon (while it's raining). We're mainly here to
>> help the people who have gone through the effort of acquiring know-how solve
>> nontrivial problems that are poorly documented or not common knowledge in
>> the community.
>> On 07/23/2014 01:23 AM, Abhishek TYAGI wrote:
>>> Dear Axel,
>>> Thankyou for your elaborated email reply, I am agree few of the points but
>>> not all, their are many users like me using NAMD and try to learn it by
>>> online resources available, and I think most of them face these problems
>>> during their work that I am facing right now. I have to all the things
>>> myself I appreciate your way of explaining each and every point in detail,
>>> this will help me to plan my work in future.
>>> regards
>>> Abhishek
>>> ________________________________________
>>> From: Axel Kohlmeyer <>
>>> Sent: Wednesday, July 23, 2014 12:36 PM
>>> To: Abhishek TYAGI
>>> Cc:
>>> Subject: Re: namd-l: Force field design
>>> On Wed, Jul 23, 2014 at 12:04 AM, Abhishek TYAGI
>>> <> wrote:
>>>> Dear Axel,
>>>> Following points I need your views:
>>>> 1. First thing is that I am willing to learn NAMD and related MD skills
>>>> for my research work, although my lab do not have subsequent knowledge about
>>>> it, but still this one of the powerful tool for research work.
>>> so attend lectures, attend tutorials, invite people to come visit and
>>> teach you or go spend some time at a group that has the necessary
>>> expertise. if you are willing you have to be willing to learn things
>>> *the right way*. at the moment, you are *not* doing this. remember
>>> that with great power comes great responsibility.
>>>> 2. If you demotivate everyone like this than in future the NAMD user
>>>> number will be effected.
>>> so what? this is not a competition about how many users a software
>>> has. that is a stupid way of seeing things. it is not then number of
>>> users that matters, but the number publications that are done
>>> correctly and properly.
>>> if you read my mail correctly, you should have seen that i do not want
>>> to demotivate you, but rather do the opposite. go to your adviser and
>>> stand up for your needs. insisting to do something without knowing
>>> what you are doing and without the proper training using an
>>> inefficient communication medium (e-mail) is something that you do to
>>> yourself and if that is demotivating and frustrating for you, then
>>> this is *your* fault, not anybody else's.
>>>> 3. NAMD should provide all the relevant solution for the users in both
>>>> the commercial and academic versions because I am not in condition to
>>>> purchase the software, that means I can't do the things, but I want to the
>>>> MDS and want to build the force field for my molecule.
>>> no. this is a ridiculous demand. please read the link that kenno
>>> provided. on top of that, your problems are not at all NAMD specific
>>> problems, but general MD problems and it is *not* the task of NAMD
>>> developers to provide the information and training how to do MD in
>>> general. why not demand this training from the charmm folks, or the
>>> amber folks, or the gromacs folks or the materials studio folks or the
>>> macromodel folks or anybody else that provides an MD code.
>>>> 4. I want to know how to do this possible task in your way, and as you
>>>> mentioned about the training I request you to please help me out for the
>>>> training thing, I will be lucky to get training under your supervision.
>>> this is the job of your adviser not mine. your adviser gets paid for
>>> it and i am not.
>>>> regards
>>>> Abhishek
>>>> ________________________________________
>>>> From: <> on behalf of
>>>> Axel Kohlmeyer <>
>>>> Sent: Wednesday, July 23, 2014 12:14 AM
>>>> To: Kenno Vanommeslaeghe
>>>> Cc:
>>>> Subject: Re: namd-l: Force field design
>>>> On Tue, Jul 22, 2014 at 11:40 AM, Kenno Vanommeslaeghe
>>>> <> wrote:
>>>>> Dear Abishek,
>>>>> In that case, you should try to make your prof read this:
>>>>> He or she basically gave you an impossible task. I don't want to
>>>>> discourage
>>>> i would not only call this an impossible task, but also outright
>>>> irresponsible behavior of the PI. not only to his student, but the
>>>> community in general. i can only encourage students to own up to this
>>>> and *demand* to be properly trained. it is not the job of the
>>>> community to compensate for this, especially when this is piled on top
>>>> of software and parameters that are provided at no cost and contain
>>>> the collected knowledge and experience of many, many people. PIs
>>>> exhibiting such irresponsible behavior should be named in public so
>>>> that a) students are made aware of the risk they may be taking and b)
>>>> their publications including simulation data, can be put under
>>>> particularly close scrutiny, since they are likely to be faulty and
>>>> misinterpreted.
>>>> axel.
>>>>> you, but it's the truth. Even if after a lot of sweat and tears, you
>>>>> somehow
>>>>> succeed to get a result out of NAMD, it will be hard to make sure you
>>>>> avoided all the nontrivial pitfalls and your results are relevant.
>>>>> Cheers,
>>>>> Kenno.
>>>>> On 07/22/2014 06:33 AM, Abhishek TYAGI wrote:
>>>>>> Dear Kenno,
>>>>>> Thanyou for your detail reply and instructions, actually I am alone
>>>>>> doing
>>>>>> work on NAMD in my lab, even in my university, my Prof do not have
>>>>>> subsequent understanding of NAMD and even I don't have so any
>>>>>> interaction to
>>>>>> Post doc/RA or other students, therefore, I directly ask you for the
>>>>>> problem. I have to do it myself and try to understand the things and do
>>>>>> them
>>>>>> carefully.
>>>>>> I hope in future you understand my situation.
>>>>>> regards
>>>>>> Abhishek
>>>>>> ________________________________________
>>>>>> From: <> on behalf of
>>>>>> Kenno Vanommeslaeghe <>
>>>>>> Sent: Monday, July 21, 2014 11:47 PM
>>>>>> To:
>>>>>> Subject: Re: namd-l: Force field design
>>>>>> Dear Abhishek,
>>>>>> This list is full of very knowledgeable people who are very good at
>>>>>> resolving people's errors, but even the greatest guru cannot help you
>>>>>> if
>>>>>> you don't tell them what exactly your problem is.
>>>>>> Since I get the impression you're relatively new to this, allow me to
>>>>>> link
>>>>>> you to CHARMM's "parameter set discussion" forum guide lines:
>>>>>> I'm NOT suggesting you ask your question there. The CHARMM forums are
>>>>>> for
>>>>>> CHARMM problems and the regulars don't like it if you post VMD/NAMD
>>>>>> problems. For NAMD, this mailinglist it the correct support channel.
>>>>>> Rather, the reason I'm giving you this link is that most of the good
>>>>>> practices it outlines are valid in *any* technical community, including
>>>>>> this list. In a nutshell, you are advised to search the web, consult
>>>>>> the
>>>>>> valid resources (tutorials, FAQs...), experiment a bit yourself and
>>>>>> consult your advisor/mentor/senior postdoc before asking a question.
>>>>>> Then,
>>>>>> if you must ask a question, take plenty of time formulating it,
>>>>>> thinking
>>>>>> of it from the point of view of the expert trying to answer it. Make it
>>>>>> concise but include all relevant information.
>>>>>> Good luck,
>>>>>> Kenno.
>>>>>> On 07/19/2014 08:44 AM, Abhishek TYAGI wrote:
>>>>>>> Dear Kenno,
>>>>>>> Thankyou for the suggestion, I had read all the information provided
>>>>>>> so
>>>>>>> far and gone through the links in detail. But still I have a silly
>>>>>>> question,
>>>>>>> right now I have a.mol2 file and I got CGenFF a.str file, and for
>>>>>>> psfgen If
>>>>>>> I use a.str file it shows error, therefore I used CHARM-GUI to get psf
>>>>>>> and
>>>>>>> pdb file, I tried doe quick md on C-GUI it gives some topology and prm
>>>>>>> file,
>>>>>>> I used them for my simulation and got error, and for other files top
>>>>>>> and prm
>>>>>>> I still confused how to proceed.
>>>>>>> regards
>>>>>>> Abhi
>>>>>>> ________________________________________
>>>>>>> From: <> on behalf of
>>>>>>> Kenno Vanommeslaeghe <>
>>>>>>> Sent: Thursday, July 17, 2014 1:24 AM
>>>>>>> To:
>>>>>>> Subject: Re: namd-l: Force field design
>>>>>>> On 07/15/2014 12:55 PM, Abhishek TYAGI wrote:
>>>>>>>> 1. I tried CGenFF and got a *.str file but the data in the file is
>>>>>>>> different from my input file i.e. mol2 file, the pdb format is
>>>>>>>> different
>>>>>>>> from the *.str file.
>>>>>>> The str file contains a force field for your molecule, while the mol2
>>>>>>> and
>>>>>>> pdb formats contain a geometry. How could the data not be different???
>>>>>>> The str file is a concatenated topology and parameter file. NAMD knows
>>>>>>> to
>>>>>>> ignore the topology section and can read the file as it is.
>>>>>>> I believe psfgen does the same (read the topology and ignore the
>>>>>>> parameters) though I'm less sure about that.
>>>>>>>> 2. I tried to get force field from swissparam website and received
>>>>>>>> the
>>>>>>>> different file as *.par, *.top file, I tried MD on this field it runs
>>>>>>>> successful, but the question is that is this correct approach?
>>>>>>> It is strongly encouraged to search the archive before posting a
>>>>>>> question.
>>>>>>> If you had, you would have discovered the use of SwissParam for
>>>>>>> condensed-phase calculations is discouraged, and found advise on
>>>>>>> better
>>>>>>> ways to do it:
>>>>>>> Christopher Mayne's advice in this thread is sound.
>>>> --
>>>> Dr. Axel Kohlmeyer
>>>> College of Science & Technology, Temple University, Philadelphia PA, USA
>>>> International Centre for Theoretical Physics, Trieste. Italy.
>>> --
>>> Dr. Axel Kohlmeyer
>>> College of Science & Technology, Temple University, Philadelphia PA, USA
>>> International Centre for Theoretical Physics, Trieste. Italy.

This archive was generated by hypermail 2.1.6 : Wed Dec 31 2014 - 23:22:39 CST