[gmx-users] LINCS warnings generated in NPT ensemble and not in NVT

[Mark Abraham]

Lee Soin wrote:
> Hello!
> I'm using COM pulling with distance constraint between two domains of a 
> protein. The energy minimization and a 100-ps equilibrating process in 
> the NPT ensemble with position restraints on protein atoms all passed 
> OK. Then I applied a distance constraint between COM of two domains and 
> performed the simulaiton in the NPT ensemble. But after the first few 
> steps I received such warnings:

"restraints" and "constraints" are different phenomena in GROMACS, and 
it's not clear from this description you're aware of that. Perhaps you 
could describe better how you implemented your effects. :-)

> Step 7  Warning: pressure scaling more than 1%, mu: 1.03819 1.03819 1.03819
> Step 8  Warning: pressure scaling more than 1%, mu: 0.98123 0.98123 0.98123
> Step 9  Warning: pressure scaling more than 1%, mu: 0.746615 0.746615 
> 0.746615
> Step 10  Warning: pressure scaling more than 1%, mu: 1.34141 1.34141 1.34141
> Step 10, time 0.02 (ps)  LINCS WARNING
> relative constraint deviation after LINCS:
> rms 0.017756, max 0.670789 (between atoms 3187 and 3188)
> bonds that rotated more than 30 degrees:
>  atom 1 atom 2  angle  previous, current, constraint length
>    4334   4336   41.3    0.0813   0.1101      0.1090
>    4334   4335   39.2    0.0813   0.1090      0.1090
>    4332   4334   32.1    0.1139   0.1548      0.1529
>    4325   4326   34.5    0.0813   0.1091      0.1090
>    4299   4301   31.1    0.1077   0.1468      0.1449
>    4293   4296   34.5    0.0814   0.1086      0.1090
>    4284   4285   30.6    0.0812   0.1093      0.1090
>    4225   4226   64.8    0.0812   0.1144      0.1090
>    4208   4210   31.0    0.1140   0.1522      0.1529
>  
> and then the program exited for too many LINCS warnings.
> The following two changes lead to no problem:
> (1) distance constraint between COM of two domains is applied, but the 
> simulation is done in the NVT ensemble;
> (2) distance restraint using an umbrella potential instead of distance 
> constraint is applied, and the simulation is done in NPT.
> So can anyone tell me what the problem is?

Maybe, if they were confident of your nomenclature.

Mark