sqm regression from AmberTools14 to 15?

[choderalab]

sqm regression from AmberTools14 to 15?

[swails]

I believe that the default convergence tolerance for the SQM calculation was decreased from AmberTools 14 to AmberTools 15. So while AmberTools 14 decided that calculation was converged, AmberTools 15 did not. You can try modifying common.c in the antechamber source code and changing the convergence criteria (making it larger) to see if it starts converging again. But a better bet might be to figure out why the molecule is not minimizing in the first place.

Also, the Amber mailing list is a better place for these kinds of questions (and make sure you're using the actual AmberTools 15 release code, not ambermini). See http://lists.ambermd.org/mailman/listinfo/amber for info about the mailing lists.

[jchodera]

@andrrizzi: Note that we should be using the canonical AM1-BCC charging scheme using OpenEye tools for all of our work, though it would be nice to be able to support pure AMBER workflows. This is implemented in get_charges() in openmoltools for now.

[jchodera]

@swails is indeed right, though. From here, Ross Walker changed this:

/* RCW Below is the old setting from AMBER 12, 14. It is ridiculously tight. 
64   * As of AmberTools 15 remove tight_p_conv and set grms_tol to 0.0005
65   * tests show it makes no major difference to the charges - max diff = 0.001.
66   *
67   * strcpy((*minfo).skeyword, "  qm_theory='AM1', grms_tol=0.0002,\n  tight_p_conv=1, scfconv=1.d-10, ");
68   *
69  */
70      strcpy((*minfo).skeyword, "  qm_theory='AM1', grms_tol=0.0005,\n  scfconv=1.d-10, ");

[andrrizzi]

I see, thanks for the pointer!

@jchodera Yes, the molecule goes through OpenEye's canonical AM1-BCC charging scheme before going to antechamber (did you mean I should use it afterwards?).

I do use epik to generate a protonation state after OpenEye though. If antechamber uses the user charges as a starting point, I could try to repeat the scheme and generate new charges to feed to antechamber. Maybe this will solve the minimization failure.

[jchodera]

@jchodera Yes, the molecule goes through OpenEye's canonical AM1-BCC charging scheme before going to antechamber (did you mean I should use it afterwards?).

If that's the case, it shouldn't run through sqm when it hits antechamber. It should just read the charges from the mol2 if no -c argument is specified (or maybe it's -c rc? but I think that is for an external charge file).

I do use epik to generate a protonation state after OpenEye though. If antechamber uses the user charges as a starting point, I could try to repeat the scheme and generate new charges to feed to antechamber. Maybe this will solve the minimization failure.

We should not be using sqm at all. The OpenEye charging should happen after epik is used to enumerate protonation states. Then these charges should be used in antechamber.

[andrrizzi]

Got it, thanks! I'll close this for now.