Continuing a FEP alchemical

[AlexanderWi]

Dear all,
I had a free energy simulation converged results harvested by alchemcial analysis in 15 lambada windows in gromacs, just to increase the lambada windows to 20, I simply added the additional lambda points to my lambda vector and ran the simulations for those in order to use TI method with out any problem. Now, I have some .xvg files with 15 columns while some of .xvg file have 20 columns. I expected that the free energy change could be calculated via TI method (alchemical analysis package) easily, so, I invoked below command to get rid of other method except TI:

"alchemical_analysis -p prd. -u kcal -f 20 -o 100dis -s 100 -c -g -v -m '-ti_cubic-dexp-iexp-bar-mbar' -i 10".
However it crashed exactly in the *.xvg file (prd.15.xvg and afterward) in which additional \lambada getting involved.

Traceback (most recent call last):
File "/usr/bin/alchemical_analysis", line 9, in
load_entry_point('alchemical-analysis==1.0.2.dev0', 'console_scripts', 'alchemical_analysis')()
File "/usr/lib/python2.7/site-packages/alchemical_analysis-1.0.2.dev0-py2.7.egg/alchemical_analysis/alchemical_analysis.py", line 1207, in main
nsnapshots, lv, dhdlt, u_klt = parser_gromacs.readDataGromacs(P)
File "/usr/lib/python2.7/site-packages/alchemical_analysis-1.0.2.dev0-py2.7.egg/alchemical_analysis/parser_gromacs.py", line 308, in readDataGromacs
nsnapshots[nf,nf] += unixlike.wcPy(f.filename) - f.skip_lines - 1*bLenConsistency
IndexError: index 15 is out of bounds for axis 1 with size 15

Any idea to sort out the problem are highly appreciated.
Regards,
Alex

[davidlmobley]

While TI can handle this easily, alchemical-analysis isn't set up to handle this case (partly because it places a heavy emphasis on BAR and MBAR, and you don't have the data for these); alchemical-analysis assumes the same number of lambda windows were used when running all of the calculations. You may be able to handle this analysis with GROMACS' built-in tools, otherwise you will be better off just processing the .xvg files yourself and extracting the data for TI analysis. It's not hard.

[halx]

On Mon, 23 Jan 2017 09:49:41 -0800 "David L. Mobley" ***@***.***> wrote: While TI can handle this easily, alchemical-analysis isn't set up to handle this case (partly because it places a heavy emphasis on BAR and MBAR, and you don't have the data for these); alchemical-analysis assumes the same number of lambda windows were used when running all of the calculations. You *may* be able to handle this analysis with GROMACS' built-in tools, otherwise you will be better off just processing the `.xvg` files yourself and extracting the data for TI analysis. It's not hard.

Since it is not hard, then alchemical-analysis should be easily able to handle it :-). Actually, I asked Alexander to report this here because I see no technical reason why this should not be possible. If the user explicitly asks not to use BAR/MBAR then the parser should deal with this by ignoring any BAR columns and deactivate any checks for those.

[davidlmobley]

@halx - I agree this is something we should probably do eventually, yes. It's just not something I can fix in the near-term (will need to wait larger architectural changes, I think). Or did you have a shorter-term fix in mind?

Thanks.

[halx]

I'm afraid I do not have an immediate short-term fix. The current design seems to be too much based on assumption as to how a _typical_ Gromacs output would look like

…

On 23 January 2017 at 18:11, David L. Mobley ***@***.***> wrote: Reopened #96 <#96>. — You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub <#96 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/AH3aWo-GBQ1Gc18f9kTYySvdivngP48Jks5rVO1JgaJpZM4LrT83> .