@Usha,

What I'd like to understand is the repeated use of the dg05 prefix of these A-Type computations. Is this indicative of what you describe in the opening post as, to sum up, a continuation of the benchmarking and re-validation cycle against a well typed reference [mentioned in past]? We all are eager to know when the real crunch commences of the 18 or so targets that were identified some 2 years ago, at start of the DDDT2 phase. See https://secure.worldcommunitygrid.org/forums/wcg/viewthread_thread,28528, where there were proximately 1000 A-Type tasks in quorum 2 were indicated per target [we just had 921 passing through], where we done 300,000+ C-type of dg05 back in January. Just so we understand and not anyone launching into [recurrent] funny notions.

thanks


In January we submitted work units for dg05 groups "a" and "b". I have submitted groups "c" and "d" now. We have grouped the 2000 ligands from the Phase 1 into group of 4 with 500 compounds each for each target dg**.

How many targets have we completely crunched through? Is dg05 the first of the 18, or have we finished any of the 18?

I know there were several "benchmarking" targets at the first, but for some reason I thought we'd done some of the real targets as well. Didn't we get some dg03 and dg04 targets at some point?

Each of dg** or ts** stands for one target. We have finished dg01, dg02, ts01-ts06 and are running dg05. dg03 and dg04 could not be finished because of the instability in the MD calculations. That is the reason we have been optimizing the code to make the code more robust. I am preparing new units ush01 and ush02 (each with 2000 ligands), which are the modified versions.

Usha,

Are we [the volunteers] to understand that the program generating the tasks is the problem and not the CHARMM program with which we are doing the computations at WCG?

--//--

These Free Energy of Binding calculations goes through the following steps:

1. Setting up the Solvent Box - we set this up in our side
2. Generalized Solvent Boundary Potential/Spherical Solvent Boundary Potential (GSBP/SSBP, which we call pre-run)
3. Grand Canonical Monte Carlo Simulations (GCMC, which we call type A) - repeated cycles of MD (10,000 steps)/MC (10,000 steps)
4. Equilibration (which we call type B) - one MD for 200,000 steps
5. Perturbations (which we call type C) - consists of ~230 separate MD simulations

The instability is not in our scripts that generates the input, it is in the parameters that go into the charmm code for these calculations mentioned above. It could be in one instance that the equilibration time for a particular system is not enough or sometimes it could be the perturbation step or sometimes the gcmc step. We have run various tests and have set up the parameters so that it might converge at every step for every calculation.

-Usha