<!DOCTYPE HTML PUBLIC "-//W3C//DTD HTML 4.01 Transitional//EN">
<html>
  <head>
    <meta content="text/html; charset=UTF-8" http-equiv="Content-Type">
  </head>
  <body text="#000000" bgcolor="#ffffff">
    On 18/01/2011 10:27 AM, TJ Mustard wrote:
    <blockquote
cite="mid:1470919845.6053.1295306828575.JavaMail.open-xchange@oxusgw00.lxa.perfora.net"
      type="cite">
      <meta content="text/html; charset=UTF-8" http-equiv="Content-Type">
      <title></title>
      <p style="margin: 0px;"><span> </span></p>
      <div style="margin: 5px 0px;"> <br>
        On January 17, 2011 at 2:47 PM TJ Mustard
        <a class="moz-txt-link-rfc2396E" href="mailto:mustardt@onid.orst.edu">&lt;mustardt@onid.orst.edu&gt;</a> wrote:<br>
        <br>
        <blockquote type="cite" style="margin-left: 0px; padding-left:
          10px; border-left: 1px solid blue;">
          <p style="margin: 0px;"><span> </span></p>
          <div style="margin: 5px 0px;"> On January 17, 2011 at 2:21 PM
            "Justin A. Lemkul" <a class="moz-txt-link-rfc2396E" href="mailto:jalemkul@vt.edu">&lt;jalemkul@vt.edu&gt;</a> wrote:<br>
            <br>
            &gt;<br>
            &gt;<br>
            &gt; TJ Mustard wrote:<br>
            &gt; &gt;<br>
            &gt; &gt;<br>
            &gt; &gt; No this is a completely different error that
            occurs when the other (EM)<br>
            &gt; &gt; error does not.<br>
            &gt; &gt;<br>
            &gt;<br>
            &gt; So, EM completes, but you then get this unstable
            system?  How well did EM<br>
            &gt; converge?  Were the potential and maximum force
            acceptable?<br>
            &gt;<br>
            &gt; Instabilities are generally caused by incorrect
            topologies or .mdp settings.<br>
            &gt; Those are the first places I would look for sources of
            error.<br>
            &gt;<br>
            &gt; -Justin<br>
            &gt;<br>
            &gt; &gt; <br>
            &gt; &gt;<br>
            &gt; &gt; Thank you,<br>
            &gt; &gt;<br>
            &gt; &gt; TJ Mustard<br>
            &gt; &gt;<br>
            &gt; &gt; On January 17, 2011 at 1:51 PM "Justin A. Lemkul"
            <a class="moz-txt-link-rfc2396E" href="mailto:jalemkul@vt.edu">&lt;jalemkul@vt.edu&gt;</a> wrote:<br>
            &gt; &gt;<br>
            &gt; &gt;  &gt;<br>
            &gt; &gt;  &gt;<br>
            &gt; &gt;  &gt; TJ Mustard wrote:<br>
            &gt; &gt;  &gt; &gt;<br>
            &gt; &gt;  &gt; &gt;<br>
            &gt; &gt;  &gt; &gt; Hi all,<br>
            &gt; &gt;  &gt; &gt;<br>
            &gt; &gt;  &gt; &gt;<br>
            &gt; &gt;  &gt; &gt;<br>
            &gt; &gt;  &gt; &gt; We have been running a lot of
            simulations on protein ligand<br>
            &gt; &gt;  &gt; &gt; interactions, and my settings/setup/mdp
            files worked great for one<br>
            &gt; &gt;  &gt; &gt; system. Then when we moved to a larger
            and more complicated system we<br>
            &gt; &gt;  &gt; &gt; started getting mdrun 1-4 interaction
            errors in our MD.<br>
            &gt; &gt;  &gt; &gt;<br>
            &gt; &gt;  &gt; &gt;<br>
            &gt; &gt;  &gt; &gt;<br>
            &gt; &gt;  &gt; &gt; We are running with heavyh and have
            correctly made our ligands with<br>
            &gt; &gt;  &gt; &gt; acpype and edited to have heavyh.<br>
            &gt; &gt;  &gt; &gt;<br>
            &gt; &gt;  &gt; &gt; Same setup as our smaller simpler
            model, but keep getting these:<br>
            &gt; &gt;  &gt; &gt;<br>
            &gt; &gt;  &gt; &gt;<br>
            &gt; &gt;  &gt; &gt;<br>
            &gt; &gt;  &gt; &gt; Warning: 1-4 interaction between X and
            Y at distance 1.804 which is<br>
            &gt; &gt;  &gt; &gt; larger than 1-4 table size 1.800 nm.<br>
            &gt; &gt;  &gt; &gt;<br>
            &gt; &gt;  &gt; &gt;<br>
            &gt; &gt;  &gt; &gt;<br>
            &gt; &gt;  &gt; &gt; The jobs will continue on and finish
            but we are worried about the error<br>
            &gt; &gt;  &gt; &gt; that could be included with these
            errors.<br>
            &gt; &gt;  &gt; &gt;<br>
            &gt; &gt;  &gt; &gt;<br>
            &gt; &gt;  &gt;<br>
            &gt; &gt;  &gt; Is this related to the other issue we're
            discussing?  If so, please<br>
            &gt; &gt; do not<br>
            &gt; &gt;  &gt; double-post, especially if you're including
            different information in<br>
            &gt; &gt; different<br>
            &gt; &gt;  &gt; threads.  It becomes a nightmare to try to
            keep track of.<br>
            &gt; &gt;  &gt;<br>
            &gt; &gt;  &gt; That particular error is usually not a good
            sign.  Normal, stable<br>
            &gt; &gt; simulations<br>
            &gt; &gt;  &gt; should not give such an error.<br>
            &gt; &gt;  &gt;<br>
            &gt; &gt;  &gt; -Justin<br>
            &gt; &gt;  &gt; </div>
          <p style="margin: 0px;"> </p>
          <p style="margin: 0px;">Here is the tail of my em:</p>
          <p style="margin: 0px;">Step= 3438, Dmax= 2.5e-04 nm, Epot=
            -1.05878e+06 Fmax= 8.91784e+01, atom= 43563<br>
            Step= 3439, Dmax= 3.0e-04 nm, Epot= -1.05878e+06 Fmax=
            8.89898e+01, atom= 43563<br>
            Step= 3440, Dmax= 3.6e-04 nm, Epot= -1.05878e+06 Fmax=
            3.27390e+02, atom= 24<br>
            Step= 3441, Dmax= 4.3e-04 nm, Epot= -1.05878e+06 Fmax=
            3.34235e+02, atom= 24<br>
            Step= 3451, Dmax= 1.0e-06 nm, Epot= -1.05878e+06 Fmax=
            3.32604e+02, atom= 24563<br>
            Stepsize too small, or no change in energy.<br>
            Converged to machine precision,<br>
            but not to the requested precision Fmax &lt; 20<br>
            <br>
            Double precision normally gives you higher accuracy.<br>
            You might need to increase your constraint accuracy, or turn<br>
            off constraints alltogether (set constraints = none in mdp
            file)<br>
            <br>
            writing lowest energy coordinates.<br>
            <br>
            Steepest Descents converged to machine precision in 3452
            steps,<br>
            but did not reach the requested Fmax &lt; 20.<br>
            Potential Energy  = -1.0587850e+06<br>
            Maximum force     =  3.3423511e+02 on atom 24<br>
            Norm of force     =  4.9464898e+00</p>
          <p style="margin: 0px;"> </p>
          <p style="margin: 0px;">My major settings are:</p>
          <p style="margin: 0px;"> </p>
          <p style="margin: 0px;">define                   =
            -DPOSRES_WATER</p>
          <p style="margin: 0px;">integrator               = steep</p>
          <p style="margin: 0px;">nsteps                   = 10000</p>
          <p style="margin: 0px;">emtol                    = 20 <br>
            emstep                   = 0.01</p>
          <p style="margin: 0px;"> </p>
          <p style="margin: 0px;">Am I setting my emtol to low? I
            occasionally see a Fmax of ~250-900 but mostly above that.</p>
          <p style="margin: 0px;"> </p>
          <p style="margin: 0px;"> </p>
          <p style="margin: 0px;">I am currently having the epiphany
            that -DPOSRES_WATER could be my problem.</p>
          <p style="margin: 0px;"> </p>
          <p style="margin: 0px;">Thank you,</p>
          <p style="margin: 0px;">TJ Mustard</p>
          <p style="margin: 0px;"> </p>
        </blockquote>
      </div>
      <p style="margin: 0px;">OK I ran the EM without the -DPOSRES_WATER
        defined and continued through my script to pr and md, and I got
        this error on my first MD after a positional restraint md (we
        call it a PR).</p>
      <p style="margin: 0px;"> </p>
      <p style="margin: 0px;">First the EM tail:</p>
      <p style="margin: 0px;"> </p>
      <p style="margin: 0px;">Step= 5519, Dmax= 8.3e-04 nm, Epot=
        -1.09261e+06 Fmax= 8.13679e+02, atom= 24<br>
        Step= 5520, Dmax= 1.0e-03 nm, Epot= -1.09261e+06 Fmax=
        7.21335e+02, atom= 24<br>
        Step= 5523, Dmax= 3.0e-04 nm, Epot= -1.09261e+06 Fmax=
        2.58642e+02, atom= 24<br>
        Step= 5527, Dmax= 4.5e-05 nm, Epot= -1.09261e+06 Fmax=
        1.88718e+02, atom= 24936<br>
        Step= 5533, Dmax= 1.7e-06 nm, Epot= -1.09261e+06 Fmax=
        1.86052e+02, atom= 24<br>
        Stepsize too small, or no change in energy.<br>
        Converged to machine precision,<br>
        but not to the requested precision Fmax &lt; 20<br>
        <br>
        Double precision normally gives you higher accuracy.<br>
        You might need to increase your constraint accuracy, or turn<br>
        off constraints alltogether (set constraints = none in mdp file)<br>
        <br>
        writing lowest energy coordinates.<br>
        <br>
        Steepest Descents converged to machine precision in 5534 steps,<br>
        but did not reach the requested Fmax &lt; 20.<br>
        Potential Energy  = -1.0926141e+06<br>
        Maximum force     =  1.8871829e+02 on atom 24<br>
        Norm of force     =  4.8353853e+00</p>
    </blockquote>
    <br>
    That looks fine.<br>
    <br>
    <blockquote
cite="mid:1470919845.6053.1295306828575.JavaMail.open-xchange@oxusgw00.lxa.perfora.net"
      type="cite">
      <p style="margin: 0px;"> </p>
      <p style="margin: 0px;">Now the MD tail:</p>
      <p style="margin: 0px;"> </p>
      <p style="margin: 0px;">No previous checkpoint file present,
        assuming this is a new run.<br>
        <br>
        Back Off! I just backed up md.log to ./#md.log.1#<br>
        Getting Loaded...<br>
        Reading file RFP-t_md.tpr, VERSION 4.5.3 (single precision)<br>
        Starting 2 threads<br>
        Loaded with Money<br>
        <br>
        Making 1D domain decomposition 2 x 1 x 1<br>
        starting mdrun 'ligand in water'<br>
        25000 steps,    100.0 ps.<br>
        step 10400, will finish Mon Jan 17 15:51:43 2011Warning: 1-4
        interaction between 108 and 116 at distance 1.800 which is
        larger than the 1-4 table size 1.800 nm<br>
        These are ignored for the rest of the simulation<br>
        This usually means your system is exploding,<br>
        if not, you should increase table-extension in your mdp file<br>
        or with user tables increase the table size<br>
        step 11700, will finish Mon Jan 17 15:52:10 2011imb F  0%</p>
      <p style="margin: 0px;"> </p>
      <p style="margin: 0px;"> </p>
      <p style="margin: 0px;">Any thoughts? Remember this system is
        doing FEP and heavy hydrogens. </p>
    </blockquote>
    <br>
    Simplify things while troubleshooting. Does normal MD work instead
    of FEP? Do normal hydrogens work?<br>
    <br>
    Use nstxout = 1 and look and the region around atoms 108 116 to see
    what is breaking. That may tell you why it is breaking.<br>
    <br>
    <blockquote
cite="mid:1470919845.6053.1295306828575.JavaMail.open-xchange@oxusgw00.lxa.perfora.net"
      type="cite">
      <p style="margin: 0px;">But so was the positional restraint md,
        and it did not error.</p>
    </blockquote>
    <br>
    Shrug, numerical instabilities can take time to show up.<br>
    <br>
    Mark<br>
  </body>
</html>