Date: Mon, 29 Jun 2020 13:46:47 -0700
Hi Daniel,
Thanks a lot for the clarification. I agree with you that the RMSF of the
pd1 and pd2 would not be the average of the individual RMSFs of pd1 and pd2.
But I was thinking that it would be somewhere close to the average value and
not greater than of the larger of the two RMSFs (for my case pd1).
I tried to play with the example you gave with a simpler system and I do see
that RMSF1234 > RMSF12 and RMSF1234 > RMSF34 when using the RMSF formula
that contains the deviation from the average value (see Case I in the
attached png).
However, when I replace this formula and compute deviations wrt a reference
value (see Case II), I do get RMSF34 > RMSF1234 > RMSF12. Since I do supply
a reference frame (first frame) I would have expected cpptraj to be doing
something like Case II.
If indeed the formula being used is similar to the Case I formula then what
is the use of the reference. Using the average position as reference (as
Kenneth suggested earlier) seems redundant.
Thanks a lot for all the help!
Kartik
From: [1]Daniel Roe
Sent: 29 June 2020 09:02
To: [2]AMBER Mailing List
Subject: Re: [AMBER] Problems with RMSF calculations
Hi,
On Fri, Jun 26, 2020 at 2:49 PM <kartik.lakshmi.gmail.com> wrote:
> I would expect the RMSF calculated using this script to give the
average
> RMSF from the two production runs
This isn't correct, and the short explanation is because the RMSF
calculation involves a square root. You can convince yourself of this
by taking a simple set of numbers (like 4 5 3 4) and plug them into
the RMSF formula ( RMSF1234 = sqrt(AvgSumSq - Avg^2), where AvgSumSq
is the average of the sum of squares of the values and Avg is the
average of the values. Then calculate RMSF for the first two and last
two values (call them RMSF12 and RMSF34). You'll see that the average
of RMSF12 and RMSF34 is not RMSF1234. Hope this helps,
-Dan
> using the input above is not the true average of the two trajectories.
>
>
> The attached graph will make it clearer.
> comp_mean.png contains the RMSF calculated using cpptraj for the entire
PD1
> and PD2 and for PD1 + PD2. I have also plotted here the mean RMSFs (for
10ns
> chunks) for PD 1 and PD 2 that I calculated using python. Am I using
the
> RMSF script incorrectly? I want to calculate the residue-wise
fluctuations
> of the Ca atoms of the peptide backbone of my entire simulation (i.e.
PD 1
> and 2 combined).
>
>
> Thanks a lot in advance for all the help!
>
> Kartik Lakshmi Rallapalli
>
> References
>
> 1.
https://urldefense.com/v3/__http:/wt_pd.nc__;!!Mih3wA!RkWVHZ9uzYaTsfRwSkKOl8
GwVSF_aGB56sRRlI2gEzyf_i_Lj7VJPgwZCZ9m$
> 2. http://pd.nc/
> 3.
https://urldefense.com/v3/__http:/wt_pd.nc__;!!Mih3wA!RkWVHZ9uzYaTsfRwSkKOl8
GwVSF_aGB56sRRlI2gEzyf_i_Lj7VJPgwZCZ9m$
> 4. http://pd.nc/
> 5.
https://urldefense.com/v3/__http:/wt_pd.nc__;!!Mih3wA!RkWVHZ9uzYaTsfRwSkKOl8
GwVSF_aGB56sRRlI2gEzyf_i_Lj7VJPgwZCZ9m$
> 6. http://pd.nc/
> _______________________________________________
> AMBER mailing list
> AMBER.ambermd.org
> http://lists.ambermd.org/mailman/listinfo/amber
_______________________________________________
AMBER mailing list
AMBER.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber
References
1. mailto:daniel.r.roe.gmail.com
2. mailto:amber.ambermd.org
_______________________________________________
AMBER mailing list
AMBER.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber
(image/png attachment: rmsf_fomula.PNG)
