search

UCL-CCS / TIES

Topology Superimposition based on joint graph traversal
MIT License
5 stars 1 forks source link

ties_ana FEP method #322

Closed Oulfin closed 1 year ago

Oulfin commented 1 year ago

Hi,

I am running some binding free energy perturbation simulations and after running simulations some transformations are not able to get the energy from the FEP method (TI works fine):

Decorrelating data... Decorrelatation was skipped using max iters per state: [3000, 3000, 3000, 3000, 3000, 3000, 3000, 3000, 3000] Failed to reach a solution to within tolerance with hybr: trying next method WARNING: Did not converge to within specified tolerance. max_delta = 0.000000e+00, tol = 1.000000e-12, maximum_iterations = 10000, iterations completed = 9999 Failed to reach a solution to within tolerance with adaptive: trying next method No solution found to within tolerance. The solution with the smallest gradient 2.828427e+03 norm is hybr Please exercise caution with this solution and consider alternative methods or a different tolerance.

pymbar.utils.ParameterError: Warning: Should have \sum_n W_nk = 1. Actual column sum for state 0 was 8.847243. 9 other columns have similar problems. This generally indicates the free energies are not converged.

My question is if is there a way to change for example tolerance (or other parameters) for the FEP method to make it work, I could not find exactly where to change (or what to change). Does it help running more replicas (now I am doing 3 replicas)? When it fails the whole process stop, is there a way to skip the directories that failed but it continues with the other directories?

Thank for the help!

Best,

Oulfin commented 1 year ago

Hi,

I could solve the issue with solver_protocol="robust" in MBAR section in FEP.py

Best,