Re: [Spam:*****] [gmx-users] REMD analysis
An acceptance ratio of 0.2/0.3 is normally best. The problem with high acceptance ratio is that it means that a large portion of the exchanges are just back and forth exchanges between consecutive exchange and are thus disturbing the system more that actually helping sampling. I do not know particularly the paper you mention but if you like what they do, it is your choice at the end. Why don;t you just increase the spacing between the replicas? You will need less replicas and potentially you could run two simulations instead of one and evaluate the convergence ... On May 16, 2013, at 1:50 AM, bharat gupta bharat.85.m...@gmail.com wrote: The plots that I showed in my last mail were for all replicas. I tried plotting the first 500 ps of replica_index and replica_time files. I think the plots look fine, and there could be problem with the plotting tool . Here the link for both files , https://www.dropbox.com/s/2g16mlxfsme4rx2/replica_temp.bmp https://www.dropbox.com/s/8jfs0b9whu6j7lo/replica_index.bmp Now regarding the high acceptance ratio which is 0.5 , I came across a paper (http://www.pnas.org/content/100/13/7587.full.pdf), here they have mentioned that their average acceptance ratio ranged between 30 to 80%. I have a question here, how did they calculate the range for the average acceptance ratio or is it average ratio for each replica . Actually, this is the reference I am following. I am also interested in peptide folding simulation, similar to this article. I want to know, whether the average acceptance ratio that I have got for my trial simulation is correct , together with the replica_temp and replica_remd plots. Can I proceed for large production runs to complete my experiment ?? On Tue, May 14, 2013 at 6:34 PM, XAvier Periole x.peri...@rug.nl wrote: The interval between the exchange trial affect the efficiency of REMD but not the the exchange ratio (at least in principle). In you case I am not sure what the plot are showing! Are these showing all the replicas? what are the units? On May 14, 2013, at 5:07 AM, bharat gupta bharat.85.m...@gmail.com wrote: Dear Sir, Here's the result for the REMD trial with large temperature gaps. Temp. distribution : 280.0 294.9 310.7 327.3 344.7 363.1 382.5 402.9 424.4 447.1 471.0 496.1 522.6 550.5 579.9 610.8 Out of md16.log : Replica exchange statistics Repl 249 attempts, 125 odd, 124 even Repl average probabilities: Repl 0123456789 10 11 12 13 14 15 Repl .40 .34 .38 .43 .43 .36 .45 .40 .37 .48 .47 .45 .47 .44 .46 Repl number of exchanges: Repl 0123456789 10 11 12 13 14 15 Repl 50 42 46 52 57 40 58 49 42 53 61 63 56 57 58 Repl average number of exchanges: Repl 0123456789 10 11 12 13 14 15 Repl .40 .34 .37 .42 .46 .32 .46 .40 .34 .43 .49 .51 .45 .46 .46 Average acceptance ratio : 0.46 But, the repli_index.xvg and replica_temp.xvg files still shows that the replicas does not exchange equally well . https://www.dropbox.com/s/zkbwpuj7l2o282b/replica_index.png https://www.dropbox.com/s/0c8gp584v1hvlbx/replica_temp.png what could be wrong in this case?? Is it the mdp file settings or implicit solvent setting. Does the time to replica to exhange also affects their swapping ?? On Tue, May 14, 2013 at 12:24 AM, XAvier Periole x.peri...@rug.nl wrote: You need to increase the temperature gaps indeed if you want acceptance ratio ~0.2/0.3. But again this won't work with the water … It is not clear what happens in your index file but probably a problem from grace to plot so many points … you can try to increase the Max drawing path length in the preference menu of grace. On May 13, 2013, at 4:22 PM, bharat gupta bharat.85.m...@gmail.com wrote: Dear Sir, I repeated the simulation again for 25 replicas with the following temp. distribution . 280 289.1 298.5 308.2 318.2 328.6 339.3 350.3 361.7 373.5 385.6 398.1 411.1 424.4 438.3 452.5 467.2 482.4 498.1 514.3 531.0 548.3 566.1 584.5 603.5 623.2 The output of md.log file is :- Replica exchange statistics Repl 24999 attempts, 12500 odd, 12499 even Repl average probabilities: Repl 0123456789 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 Repl .63 .63 .62 .62 .61 .61 .60 .60 .59 .59 .58 .59 .59 .60 .60 .61 .62 .62 .63 .64 .64 .65 .65 .66 .66 Repl number of exchanges: Repl 0123456789 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 Repl 7822 7752 7816 7760 7639 7628 7511 7442 7375 7332 7312 7424 7408 7410 7522 7559 7684 7697 7878 7927 7917 8073 8151 8208 8266
Re: [Spam:*****] [gmx-users] REMD analysis
Okay Sir, I will try two-three combinations this time and will report back to you ... On Thu, May 16, 2013 at 5:25 PM, XAvier Periole x.peri...@rug.nl wrote: An acceptance ratio of 0.2/0.3 is normally best. The problem with high acceptance ratio is that it means that a large portion of the exchanges are just back and forth exchanges between consecutive exchange and are thus disturbing the system more that actually helping sampling. I do not know particularly the paper you mention but if you like what they do, it is your choice at the end. Why don;t you just increase the spacing between the replicas? You will need less replicas and potentially you could run two simulations instead of one and evaluate the convergence ... On May 16, 2013, at 1:50 AM, bharat gupta bharat.85.m...@gmail.com wrote: The plots that I showed in my last mail were for all replicas. I tried plotting the first 500 ps of replica_index and replica_time files. I think the plots look fine, and there could be problem with the plotting tool . Here the link for both files , https://www.dropbox.com/s/2g16mlxfsme4rx2/replica_temp.bmp https://www.dropbox.com/s/8jfs0b9whu6j7lo/replica_index.bmp Now regarding the high acceptance ratio which is 0.5 , I came across a paper (http://www.pnas.org/content/100/13/7587.full.pdf), here they have mentioned that their average acceptance ratio ranged between 30 to 80%. I have a question here, how did they calculate the range for the average acceptance ratio or is it average ratio for each replica . Actually, this is the reference I am following. I am also interested in peptide folding simulation, similar to this article. I want to know, whether the average acceptance ratio that I have got for my trial simulation is correct , together with the replica_temp and replica_remd plots. Can I proceed for large production runs to complete my experiment ?? On Tue, May 14, 2013 at 6:34 PM, XAvier Periole x.peri...@rug.nl wrote: The interval between the exchange trial affect the efficiency of REMD but not the the exchange ratio (at least in principle). In you case I am not sure what the plot are showing! Are these showing all the replicas? what are the units? On May 14, 2013, at 5:07 AM, bharat gupta bharat.85.m...@gmail.com wrote: Dear Sir, Here's the result for the REMD trial with large temperature gaps. Temp. distribution : 280.0 294.9 310.7 327.3 344.7 363.1 382.5 402.9 424.4 447.1 471.0 496.1 522.6 550.5 579.9 610.8 Out of md16.log : Replica exchange statistics Repl 249 attempts, 125 odd, 124 even Repl average probabilities: Repl 0123456789 10 11 12 13 14 15 Repl .40 .34 .38 .43 .43 .36 .45 .40 .37 .48 .47 .45 .47 .44 .46 Repl number of exchanges: Repl 0123456789 10 11 12 13 14 15 Repl 50 42 46 52 57 40 58 49 42 53 61 63 56 57 58 Repl average number of exchanges: Repl 0123456789 10 11 12 13 14 15 Repl .40 .34 .37 .42 .46 .32 .46 .40 .34 .43 .49 .51 .45 .46 .46 Average acceptance ratio : 0.46 But, the repli_index.xvg and replica_temp.xvg files still shows that the replicas does not exchange equally well . https://www.dropbox.com/s/zkbwpuj7l2o282b/replica_index.png https://www.dropbox.com/s/0c8gp584v1hvlbx/replica_temp.png what could be wrong in this case?? Is it the mdp file settings or implicit solvent setting. Does the time to replica to exhange also affects their swapping ?? On Tue, May 14, 2013 at 12:24 AM, XAvier Periole x.peri...@rug.nl wrote: You need to increase the temperature gaps indeed if you want acceptance ratio ~0.2/0.3. But again this won't work with the water … It is not clear what happens in your index file but probably a problem from grace to plot so many points … you can try to increase the Max drawing path length in the preference menu of grace. On May 13, 2013, at 4:22 PM, bharat gupta bharat.85.m...@gmail.com wrote: Dear Sir, I repeated the simulation again for 25 replicas with the following temp. distribution . 280 289.1 298.5 308.2 318.2 328.6 339.3 350.3 361.7 373.5 385.6 398.1 411.1 424.4 438.3 452.5 467.2 482.4 498.1 514.3 531.0 548.3 566.1 584.5 603.5 623.2 The output of md.log file is :- Replica exchange statistics Repl 24999 attempts, 12500 odd, 12499 even Repl average probabilities: Repl 0123456789 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 Repl .63 .63 .62 .62 .61 .61 .60 .60 .59 .59 .58 .59 .59 .60 .60 .61 .62 .62 .63 .64 .64 .65 .65 .66 .66 Repl number of
Re: [Spam:*****] [gmx-users] REMD analysis
You have to convince yourself, not me :)) But I can give you my opinion … On May 16, 2013, at 10:33 AM, bharat gupta bharat.85.m...@gmail.com wrote: Okay Sir, I will try two-three combinations this time and will report back to you ... On Thu, May 16, 2013 at 5:25 PM, XAvier Periole x.peri...@rug.nl wrote: An acceptance ratio of 0.2/0.3 is normally best. The problem with high acceptance ratio is that it means that a large portion of the exchanges are just back and forth exchanges between consecutive exchange and are thus disturbing the system more that actually helping sampling. I do not know particularly the paper you mention but if you like what they do, it is your choice at the end. Why don;t you just increase the spacing between the replicas? You will need less replicas and potentially you could run two simulations instead of one and evaluate the convergence ... On May 16, 2013, at 1:50 AM, bharat gupta bharat.85.m...@gmail.com wrote: The plots that I showed in my last mail were for all replicas. I tried plotting the first 500 ps of replica_index and replica_time files. I think the plots look fine, and there could be problem with the plotting tool . Here the link for both files , https://www.dropbox.com/s/2g16mlxfsme4rx2/replica_temp.bmp https://www.dropbox.com/s/8jfs0b9whu6j7lo/replica_index.bmp Now regarding the high acceptance ratio which is 0.5 , I came across a paper (http://www.pnas.org/content/100/13/7587.full.pdf), here they have mentioned that their average acceptance ratio ranged between 30 to 80%. I have a question here, how did they calculate the range for the average acceptance ratio or is it average ratio for each replica . Actually, this is the reference I am following. I am also interested in peptide folding simulation, similar to this article. I want to know, whether the average acceptance ratio that I have got for my trial simulation is correct , together with the replica_temp and replica_remd plots. Can I proceed for large production runs to complete my experiment ?? On Tue, May 14, 2013 at 6:34 PM, XAvier Periole x.peri...@rug.nl wrote: The interval between the exchange trial affect the efficiency of REMD but not the the exchange ratio (at least in principle). In you case I am not sure what the plot are showing! Are these showing all the replicas? what are the units? On May 14, 2013, at 5:07 AM, bharat gupta bharat.85.m...@gmail.com wrote: Dear Sir, Here's the result for the REMD trial with large temperature gaps. Temp. distribution : 280.0 294.9 310.7 327.3 344.7 363.1 382.5 402.9 424.4 447.1 471.0 496.1 522.6 550.5 579.9 610.8 Out of md16.log : Replica exchange statistics Repl 249 attempts, 125 odd, 124 even Repl average probabilities: Repl 0123456789 10 11 12 13 14 15 Repl .40 .34 .38 .43 .43 .36 .45 .40 .37 .48 .47 .45 .47 .44 .46 Repl number of exchanges: Repl 0123456789 10 11 12 13 14 15 Repl 50 42 46 52 57 40 58 49 42 53 61 63 56 57 58 Repl average number of exchanges: Repl 0123456789 10 11 12 13 14 15 Repl .40 .34 .37 .42 .46 .32 .46 .40 .34 .43 .49 .51 .45 .46 .46 Average acceptance ratio : 0.46 But, the repli_index.xvg and replica_temp.xvg files still shows that the replicas does not exchange equally well . https://www.dropbox.com/s/zkbwpuj7l2o282b/replica_index.png https://www.dropbox.com/s/0c8gp584v1hvlbx/replica_temp.png what could be wrong in this case?? Is it the mdp file settings or implicit solvent setting. Does the time to replica to exhange also affects their swapping ?? On Tue, May 14, 2013 at 12:24 AM, XAvier Periole x.peri...@rug.nl wrote: You need to increase the temperature gaps indeed if you want acceptance ratio ~0.2/0.3. But again this won't work with the water … It is not clear what happens in your index file but probably a problem from grace to plot so many points … you can try to increase the Max drawing path length in the preference menu of grace. On May 13, 2013, at 4:22 PM, bharat gupta bharat.85.m...@gmail.com wrote: Dear Sir, I repeated the simulation again for 25 replicas with the following temp. distribution . 280 289.1 298.5 308.2 318.2 328.6 339.3 350.3 361.7 373.5 385.6 398.1 411.1 424.4 438.3 452.5 467.2 482.4 498.1 514.3 531.0 548.3 566.1 584.5 603.5 623.2 The output of md.log file is :- Replica exchange statistics Repl 24999 attempts, 12500 odd, 12499 even Repl average probabilities: Repl 0123456789 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 Repl .63 .63 .62 .62 .61 .61 .60 .60 .59 .59 .58 .59 .59 .60 .60 .61 .62 .62 .63
Re: [Spam:*****] [gmx-users] REMD analysis
Dear Sir, Here's the result of three different runs : Temperature distribution for three trials Repeat-1 280 298 317 337 359 382 406 432 460 489 520 554 589 627 Repeat-2 280 299 319 340 363 388 414 441 471 503 536 572 611 Repeat-3 280 300 322 345 370 397 426 457 490 526 564 605 649 md.log files output from three different trials: Repeat-1 .37 .28 .26 .30 .25 .29 .32 .35 .32 .35 .36 .32 .31 Repeat-2 .30 .33 .30 .25 .19 .27 .30 .31 .27 .40 .34 .31 Repeat-3 .18 .22 .26 .34 .26 .28 .25 .27 .27 .25 .27 .22 I think as the required acceptance value all the three trials are fine, but trail 3 would be much better to continue the further runs and anlysis ?? So, is it fine to continue with the third simulation ?? But still the problem is that I am not getting the exact graphs with xmgrace?? On Thu, May 16, 2013 at 5:36 PM, XAvier Periole x.peri...@rug.nl wrote: You have to convince yourself, not me :)) But I can give you my opinion … On May 16, 2013, at 10:33 AM, bharat gupta bharat.85.m...@gmail.com wrote: Okay Sir, I will try two-three combinations this time and will report back to you ... On Thu, May 16, 2013 at 5:25 PM, XAvier Periole x.peri...@rug.nl wrote: An acceptance ratio of 0.2/0.3 is normally best. The problem with high acceptance ratio is that it means that a large portion of the exchanges are just back and forth exchanges between consecutive exchange and are thus disturbing the system more that actually helping sampling. I do not know particularly the paper you mention but if you like what they do, it is your choice at the end. Why don;t you just increase the spacing between the replicas? You will need less replicas and potentially you could run two simulations instead of one and evaluate the convergence ... On May 16, 2013, at 1:50 AM, bharat gupta bharat.85.m...@gmail.com wrote: The plots that I showed in my last mail were for all replicas. I tried plotting the first 500 ps of replica_index and replica_time files. I think the plots look fine, and there could be problem with the plotting tool . Here the link for both files , https://www.dropbox.com/s/2g16mlxfsme4rx2/replica_temp.bmp https://www.dropbox.com/s/8jfs0b9whu6j7lo/replica_index.bmp Now regarding the high acceptance ratio which is 0.5 , I came across a paper (http://www.pnas.org/content/100/13/7587.full.pdf), here they have mentioned that their average acceptance ratio ranged between 30 to 80%. I have a question here, how did they calculate the range for the average acceptance ratio or is it average ratio for each replica . Actually, this is the reference I am following. I am also interested in peptide folding simulation, similar to this article. I want to know, whether the average acceptance ratio that I have got for my trial simulation is correct , together with the replica_temp and replica_remd plots. Can I proceed for large production runs to complete my experiment ?? On Tue, May 14, 2013 at 6:34 PM, XAvier Periole x.peri...@rug.nl wrote: The interval between the exchange trial affect the efficiency of REMD but not the the exchange ratio (at least in principle). In you case I am not sure what the plot are showing! Are these showing all the replicas? what are the units? On May 14, 2013, at 5:07 AM, bharat gupta bharat.85.m...@gmail.com wrote: Dear Sir, Here's the result for the REMD trial with large temperature gaps. Temp. distribution : 280.0 294.9 310.7 327.3 344.7 363.1 382.5 402.9 424.4 447.1 471.0 496.1 522.6 550.5 579.9 610.8 Out of md16.log : Replica exchange statistics Repl 249 attempts, 125 odd, 124 even Repl average probabilities: Repl 0123456789 10 11 12 13 14 15 Repl .40 .34 .38 .43 .43 .36 .45 .40 .37 .48 .47 .45 .47 .44 .46 Repl number of exchanges: Repl 0123456789 10 11 12 13 14 15 Repl 50 42 46 52 57 40 58 49 42 53 61 63 56 57 58 Repl average number of exchanges: Repl 0123456789 10 11 12 13 14 15 Repl .40 .34 .37 .42 .46 .32 .46 .40 .34 .43 .49 .51 .45 .46 .46 Average acceptance ratio : 0.46 But, the repli_index.xvg and replica_temp.xvg files still shows that the replicas does not exchange equally well . https://www.dropbox.com/s/zkbwpuj7l2o282b/replica_index.png https://www.dropbox.com/s/0c8gp584v1hvlbx/replica_temp.png what could be wrong in this case?? Is it the mdp file settings or implicit solvent setting. Does the time to replica to exhange also affects their swapping ?? On Tue, May 14, 2013 at 12:24 AM, XAvier Periole x.peri...@rug.nl wrote: You need to increase the temperature gaps indeed if you want
Re: [Spam:*****] [gmx-users] REMD analysis
On Thu, May 16, 2013 at 2:04 PM, bharat gupta bharat.85.m...@gmail.comwrote: Dear Sir, Here's the result of three different runs : Temperature distribution for three trials Repeat-1 280 298 317 337 359 382 406 432 460 489 520 554 589 627 Repeat-2 280 299 319 340 363 388 414 441 471 503 536 572 611 Repeat-3 280 300 322 345 370 397 426 457 490 526 564 605 649 md.log files output from three different trials: Repeat-1 .37 .28 .26 .30 .25 .29 .32 .35 .32 .35 .36 .32 .31 Repeat-2 .30 .33 .30 .25 .19 .27 .30 .31 .27 .40 .34 .31 Repeat-3 .18 .22 .26 .34 .26 .28 .25 .27 .27 .25 .27 .22 I think as the required acceptance value all the three trials are fine, but trail 3 would be much better to continue the further runs and anlysis ?? Probably. But exchange acceptance is a poor proxy for sampling efficiency - see recent discussions of REMD on this list. Mark So, is it fine to continue with the third simulation ?? But still the problem is that I am not getting the exact graphs with xmgrace?? On Thu, May 16, 2013 at 5:36 PM, XAvier Periole x.peri...@rug.nl wrote: You have to convince yourself, not me :)) But I can give you my opinion … On May 16, 2013, at 10:33 AM, bharat gupta bharat.85.m...@gmail.com wrote: Okay Sir, I will try two-three combinations this time and will report back to you ... On Thu, May 16, 2013 at 5:25 PM, XAvier Periole x.peri...@rug.nl wrote: An acceptance ratio of 0.2/0.3 is normally best. The problem with high acceptance ratio is that it means that a large portion of the exchanges are just back and forth exchanges between consecutive exchange and are thus disturbing the system more that actually helping sampling. I do not know particularly the paper you mention but if you like what they do, it is your choice at the end. Why don;t you just increase the spacing between the replicas? You will need less replicas and potentially you could run two simulations instead of one and evaluate the convergence ... On May 16, 2013, at 1:50 AM, bharat gupta bharat.85.m...@gmail.com wrote: The plots that I showed in my last mail were for all replicas. I tried plotting the first 500 ps of replica_index and replica_time files. I think the plots look fine, and there could be problem with the plotting tool . Here the link for both files , https://www.dropbox.com/s/2g16mlxfsme4rx2/replica_temp.bmp https://www.dropbox.com/s/8jfs0b9whu6j7lo/replica_index.bmp Now regarding the high acceptance ratio which is 0.5 , I came across a paper (http://www.pnas.org/content/100/13/7587.full.pdf), here they have mentioned that their average acceptance ratio ranged between 30 to 80%. I have a question here, how did they calculate the range for the average acceptance ratio or is it average ratio for each replica . Actually, this is the reference I am following. I am also interested in peptide folding simulation, similar to this article. I want to know, whether the average acceptance ratio that I have got for my trial simulation is correct , together with the replica_temp and replica_remd plots. Can I proceed for large production runs to complete my experiment ?? On Tue, May 14, 2013 at 6:34 PM, XAvier Periole x.peri...@rug.nl wrote: The interval between the exchange trial affect the efficiency of REMD but not the the exchange ratio (at least in principle). In you case I am not sure what the plot are showing! Are these showing all the replicas? what are the units? On May 14, 2013, at 5:07 AM, bharat gupta bharat.85.m...@gmail.com wrote: Dear Sir, Here's the result for the REMD trial with large temperature gaps. Temp. distribution : 280.0 294.9 310.7 327.3 344.7 363.1 382.5 402.9 424.4 447.1 471.0 496.1 522.6 550.5 579.9 610.8 Out of md16.log : Replica exchange statistics Repl 249 attempts, 125 odd, 124 even Repl average probabilities: Repl 0123456789 10 11 12 13 14 15 Repl .40 .34 .38 .43 .43 .36 .45 .40 .37 .48 .47 .45 .47 .44 .46 Repl number of exchanges: Repl 0123456789 10 11 12 13 14 15 Repl 50 42 46 52 57 40 58 49 42 53 61 63 56 57 58 Repl average number of exchanges: Repl 0123456789 10 11 12 13 14 15 Repl .40 .34 .37 .42 .46 .32 .46 .40 .34 .43 .49 .51 .45 .46 .46 Average acceptance ratio : 0.46 But, the repli_index.xvg and replica_temp.xvg files still shows that the replicas does not exchange equally well . https://www.dropbox.com/s/zkbwpuj7l2o282b/replica_index.png
Re: [Spam:*****] [gmx-users] REMD analysis
Sorry to ask this simple question but how to read the replica_index and replica_temp files. I tried to search a lot but didn't find any information. As I have concatenated all log files and demuxed them. Here's first 10 lines from both files:- replica_index: 0 0123456789 10 11 12 2 1023456798 10 11 12 4 1023456978 10 11 12 6 1023459678 10 11 12 8 1203495768 10 11 12 10 1230947568 11 10 12 replica_temp 0 0123456789 10 11 12 2 1023456798 10 11 12 4 1023456897 10 11 12 6 1023457896 10 11 12 8 2013468795 10 11 12 10 3012578694 11 10 12 On Thu, May 16, 2013 at 11:24 PM, Mark Abraham mark.j.abra...@gmail.comwrote: On Thu, May 16, 2013 at 2:04 PM, bharat gupta bharat.85.m...@gmail.com wrote: Dear Sir, Here's the result of three different runs : Temperature distribution for three trials Repeat-1 280 298 317 337 359 382 406 432 460 489 520 554 589 627 Repeat-2 280 299 319 340 363 388 414 441 471 503 536 572 611 Repeat-3 280 300 322 345 370 397 426 457 490 526 564 605 649 md.log files output from three different trials: Repeat-1 .37 .28 .26 .30 .25 .29 .32 .35 .32 .35 .36 .32 .31 Repeat-2 .30 .33 .30 .25 .19 .27 .30 .31 .27 .40 .34 .31 Repeat-3 .18 .22 .26 .34 .26 .28 .25 .27 .27 .25 .27 .22 I think as the required acceptance value all the three trials are fine, but trail 3 would be much better to continue the further runs and anlysis ?? Probably. But exchange acceptance is a poor proxy for sampling efficiency - see recent discussions of REMD on this list. Mark So, is it fine to continue with the third simulation ?? But still the problem is that I am not getting the exact graphs with xmgrace?? On Thu, May 16, 2013 at 5:36 PM, XAvier Periole x.peri...@rug.nl wrote: You have to convince yourself, not me :)) But I can give you my opinion … On May 16, 2013, at 10:33 AM, bharat gupta bharat.85.m...@gmail.com wrote: Okay Sir, I will try two-three combinations this time and will report back to you ... On Thu, May 16, 2013 at 5:25 PM, XAvier Periole x.peri...@rug.nl wrote: An acceptance ratio of 0.2/0.3 is normally best. The problem with high acceptance ratio is that it means that a large portion of the exchanges are just back and forth exchanges between consecutive exchange and are thus disturbing the system more that actually helping sampling. I do not know particularly the paper you mention but if you like what they do, it is your choice at the end. Why don;t you just increase the spacing between the replicas? You will need less replicas and potentially you could run two simulations instead of one and evaluate the convergence ... On May 16, 2013, at 1:50 AM, bharat gupta bharat.85.m...@gmail.com wrote: The plots that I showed in my last mail were for all replicas. I tried plotting the first 500 ps of replica_index and replica_time files. I think the plots look fine, and there could be problem with the plotting tool . Here the link for both files , https://www.dropbox.com/s/2g16mlxfsme4rx2/replica_temp.bmp https://www.dropbox.com/s/8jfs0b9whu6j7lo/replica_index.bmp Now regarding the high acceptance ratio which is 0.5 , I came across a paper (http://www.pnas.org/content/100/13/7587.full.pdf), here they have mentioned that their average acceptance ratio ranged between 30 to 80%. I have a question here, how did they calculate the range for the average acceptance ratio or is it average ratio for each replica . Actually, this is the reference I am following. I am also interested in peptide folding simulation, similar to this article. I want to know, whether the average acceptance ratio that I have got for my trial simulation is correct , together with the replica_temp and replica_remd plots. Can I proceed for large production runs to complete my experiment ?? On Tue, May 14, 2013 at 6:34 PM, XAvier Periole x.peri...@rug.nl wrote: The interval between the exchange trial affect the efficiency of REMD but not the the exchange ratio (at least in principle). In you case I am not sure what the plot are showing! Are
Re: [Spam:*****] [gmx-users] REMD analysis
They show which structure is in which ensemble, and the inverse. Look at the exchange events reported in the .log files and work out which is which. Mark On Thu, May 16, 2013 at 5:25 PM, bharat gupta bharat.85.m...@gmail.comwrote: Sorry to ask this simple question but how to read the replica_index and replica_temp files. I tried to search a lot but didn't find any information. As I have concatenated all log files and demuxed them. Here's first 10 lines from both files:- replica_index: 0 0123456789 10 11 12 2 1023456798 10 11 12 4 1023456978 10 11 12 6 1023459678 10 11 12 8 1203495768 10 11 12 10 1230947568 11 10 12 replica_temp 0 0123456789 10 11 12 2 1023456798 10 11 12 4 1023456897 10 11 12 6 1023457896 10 11 12 8 2013468795 10 11 12 10 3012578694 11 10 12 On Thu, May 16, 2013 at 11:24 PM, Mark Abraham mark.j.abra...@gmail.com wrote: On Thu, May 16, 2013 at 2:04 PM, bharat gupta bharat.85.m...@gmail.com wrote: Dear Sir, Here's the result of three different runs : Temperature distribution for three trials Repeat-1 280 298 317 337 359 382 406 432 460 489 520 554 589 627 Repeat-2 280 299 319 340 363 388 414 441 471 503 536 572 611 Repeat-3 280 300 322 345 370 397 426 457 490 526 564 605 649 md.log files output from three different trials: Repeat-1 .37 .28 .26 .30 .25 .29 .32 .35 .32 .35 .36 .32 .31 Repeat-2 .30 .33 .30 .25 .19 .27 .30 .31 .27 .40 .34 .31 Repeat-3 .18 .22 .26 .34 .26 .28 .25 .27 .27 .25 .27 .22 I think as the required acceptance value all the three trials are fine, but trail 3 would be much better to continue the further runs and anlysis ?? Probably. But exchange acceptance is a poor proxy for sampling efficiency - see recent discussions of REMD on this list. Mark So, is it fine to continue with the third simulation ?? But still the problem is that I am not getting the exact graphs with xmgrace?? On Thu, May 16, 2013 at 5:36 PM, XAvier Periole x.peri...@rug.nl wrote: You have to convince yourself, not me :)) But I can give you my opinion … On May 16, 2013, at 10:33 AM, bharat gupta bharat.85.m...@gmail.com wrote: Okay Sir, I will try two-three combinations this time and will report back to you ... On Thu, May 16, 2013 at 5:25 PM, XAvier Periole x.peri...@rug.nl wrote: An acceptance ratio of 0.2/0.3 is normally best. The problem with high acceptance ratio is that it means that a large portion of the exchanges are just back and forth exchanges between consecutive exchange and are thus disturbing the system more that actually helping sampling. I do not know particularly the paper you mention but if you like what they do, it is your choice at the end. Why don;t you just increase the spacing between the replicas? You will need less replicas and potentially you could run two simulations instead of one and evaluate the convergence ... On May 16, 2013, at 1:50 AM, bharat gupta bharat.85.m...@gmail.com wrote: The plots that I showed in my last mail were for all replicas. I tried plotting the first 500 ps of replica_index and replica_time files. I think the plots look fine, and there could be problem with the plotting tool . Here the link for both files , https://www.dropbox.com/s/2g16mlxfsme4rx2/replica_temp.bmp https://www.dropbox.com/s/8jfs0b9whu6j7lo/replica_index.bmp Now regarding the high acceptance ratio which is 0.5 , I came across a paper (http://www.pnas.org/content/100/13/7587.full.pdf), here they have mentioned that their average acceptance ratio ranged between 30 to 80%. I have a question here, how did they calculate the range for the average acceptance ratio or is it average ratio for each replica . Actually, this is the reference I am following. I am also interested in peptide folding simulation, similar to this article. I want to know, whether the average acceptance ratio that I have got for my trial simulation is correct , together with the replica_temp and replica_remd plots. Can I
Re: [Spam:*****] [gmx-users] REMD analysis
The plots that I showed in my last mail were for all replicas. I tried plotting the first 500 ps of replica_index and replica_time files. I think the plots look fine, and there could be problem with the plotting tool . Here the link for both files , https://www.dropbox.com/s/2g16mlxfsme4rx2/replica_temp.bmp https://www.dropbox.com/s/8jfs0b9whu6j7lo/replica_index.bmp Now regarding the high acceptance ratio which is 0.5 , I came across a paper (http://www.pnas.org/content/100/13/7587.full.pdf), here they have mentioned that their average acceptance ratio ranged between 30 to 80%. I have a question here, how did they calculate the range for the average acceptance ratio or is it average ratio for each replica . Actually, this is the reference I am following. I am also interested in peptide folding simulation, similar to this article. I want to know, whether the average acceptance ratio that I have got for my trial simulation is correct , together with the replica_temp and replica_remd plots. Can I proceed for large production runs to complete my experiment ?? On Tue, May 14, 2013 at 6:34 PM, XAvier Periole x.peri...@rug.nl wrote: The interval between the exchange trial affect the efficiency of REMD but not the the exchange ratio (at least in principle). In you case I am not sure what the plot are showing! Are these showing all the replicas? what are the units? On May 14, 2013, at 5:07 AM, bharat gupta bharat.85.m...@gmail.com wrote: Dear Sir, Here's the result for the REMD trial with large temperature gaps. Temp. distribution : 280.0 294.9 310.7 327.3 344.7 363.1 382.5 402.9 424.4 447.1 471.0 496.1 522.6 550.5 579.9 610.8 Out of md16.log : Replica exchange statistics Repl 249 attempts, 125 odd, 124 even Repl average probabilities: Repl 0123456789 10 11 12 13 14 15 Repl .40 .34 .38 .43 .43 .36 .45 .40 .37 .48 .47 .45 .47 .44 .46 Repl number of exchanges: Repl 0123456789 10 11 12 13 14 15 Repl 50 42 46 52 57 40 58 49 42 53 61 63 56 57 58 Repl average number of exchanges: Repl 0123456789 10 11 12 13 14 15 Repl .40 .34 .37 .42 .46 .32 .46 .40 .34 .43 .49 .51 .45 .46 .46 Average acceptance ratio : 0.46 But, the repli_index.xvg and replica_temp.xvg files still shows that the replicas does not exchange equally well . https://www.dropbox.com/s/zkbwpuj7l2o282b/replica_index.png https://www.dropbox.com/s/0c8gp584v1hvlbx/replica_temp.png what could be wrong in this case?? Is it the mdp file settings or implicit solvent setting. Does the time to replica to exhange also affects their swapping ?? On Tue, May 14, 2013 at 12:24 AM, XAvier Periole x.peri...@rug.nl wrote: You need to increase the temperature gaps indeed if you want acceptance ratio ~0.2/0.3. But again this won't work with the water … It is not clear what happens in your index file but probably a problem from grace to plot so many points … you can try to increase the Max drawing path length in the preference menu of grace. On May 13, 2013, at 4:22 PM, bharat gupta bharat.85.m...@gmail.com wrote: Dear Sir, I repeated the simulation again for 25 replicas with the following temp. distribution . 280 289.1 298.5 308.2 318.2 328.6 339.3 350.3 361.7 373.5 385.6 398.1 411.1 424.4 438.3 452.5 467.2 482.4 498.1 514.3 531.0 548.3 566.1 584.5 603.5 623.2 The output of md.log file is :- Replica exchange statistics Repl 24999 attempts, 12500 odd, 12499 even Repl average probabilities: Repl 0123456789 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 Repl .63 .63 .62 .62 .61 .61 .60 .60 .59 .59 .58 .59 .59 .60 .60 .61 .62 .62 .63 .64 .64 .65 .65 .66 .66 Repl number of exchanges: Repl 0123456789 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 Repl 7822 7752 7816 7760 7639 7628 7511 7442 7375 7332 7312 7424 7408 7410 7522 7559 7684 7697 7878 7927 7917 8073 8151 8208 8266 Repl average number of exchanges: Repl 0123456789 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 Repl .63 .62 .63 .62 .61 .61 .60 .60 .59 .59 .58 .59 .59 .59 .60 .60 .61 .62 .63 .63 .63 .65 .65 .66 .66 The average acceptance ration is around 0.6 which is still high. The link for replica_temp,replica_index : https://www.dropbox.com/s/c7soajnwc3uww8j/replica_temp.png https://www.dropbox.com/s/wvx82m4c6cnsfit/replica_index.png The temp files look better but the index file looks
Re: [Spam:*****] [gmx-users] REMD analysis
The interval between the exchange trial affect the efficiency of REMD but not the the exchange ratio (at least in principle). In you case I am not sure what the plot are showing! Are these showing all the replicas? what are the units? On May 14, 2013, at 5:07 AM, bharat gupta bharat.85.m...@gmail.com wrote: Dear Sir, Here's the result for the REMD trial with large temperature gaps. Temp. distribution : 280.0 294.9 310.7 327.3 344.7 363.1 382.5 402.9 424.4 447.1 471.0 496.1 522.6 550.5 579.9 610.8 Out of md16.log : Replica exchange statistics Repl 249 attempts, 125 odd, 124 even Repl average probabilities: Repl 0123456789 10 11 12 13 14 15 Repl .40 .34 .38 .43 .43 .36 .45 .40 .37 .48 .47 .45 .47 .44 .46 Repl number of exchanges: Repl 0123456789 10 11 12 13 14 15 Repl 50 42 46 52 57 40 58 49 42 53 61 63 56 57 58 Repl average number of exchanges: Repl 0123456789 10 11 12 13 14 15 Repl .40 .34 .37 .42 .46 .32 .46 .40 .34 .43 .49 .51 .45 .46 .46 Average acceptance ratio : 0.46 But, the repli_index.xvg and replica_temp.xvg files still shows that the replicas does not exchange equally well . https://www.dropbox.com/s/zkbwpuj7l2o282b/replica_index.png https://www.dropbox.com/s/0c8gp584v1hvlbx/replica_temp.png what could be wrong in this case?? Is it the mdp file settings or implicit solvent setting. Does the time to replica to exhange also affects their swapping ?? On Tue, May 14, 2013 at 12:24 AM, XAvier Periole x.peri...@rug.nl wrote: You need to increase the temperature gaps indeed if you want acceptance ratio ~0.2/0.3. But again this won't work with the water … It is not clear what happens in your index file but probably a problem from grace to plot so many points … you can try to increase the Max drawing path length in the preference menu of grace. On May 13, 2013, at 4:22 PM, bharat gupta bharat.85.m...@gmail.com wrote: Dear Sir, I repeated the simulation again for 25 replicas with the following temp. distribution . 280 289.1 298.5 308.2 318.2 328.6 339.3 350.3 361.7 373.5 385.6 398.1 411.1 424.4 438.3 452.5 467.2 482.4 498.1 514.3 531.0 548.3 566.1 584.5 603.5 623.2 The output of md.log file is :- Replica exchange statistics Repl 24999 attempts, 12500 odd, 12499 even Repl average probabilities: Repl 0123456789 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 Repl .63 .63 .62 .62 .61 .61 .60 .60 .59 .59 .58 .59 .59 .60 .60 .61 .62 .62 .63 .64 .64 .65 .65 .66 .66 Repl number of exchanges: Repl 0123456789 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 Repl 7822 7752 7816 7760 7639 7628 7511 7442 7375 7332 7312 7424 7408 7410 7522 7559 7684 7697 7878 7927 7917 8073 8151 8208 8266 Repl average number of exchanges: Repl 0123456789 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 Repl .63 .62 .63 .62 .61 .61 .60 .60 .59 .59 .58 .59 .59 .59 .60 .60 .61 .62 .63 .63 .63 .65 .65 .66 .66 The average acceptance ration is around 0.6 which is still high. The link for replica_temp,replica_index : https://www.dropbox.com/s/c7soajnwc3uww8j/replica_temp.png https://www.dropbox.com/s/wvx82m4c6cnsfit/replica_index.png The temp files look better but the index file looks weird ... Do i need to experiment with the gap difference in order to get the required ration of 0.2-0.3 ?? There is some problem with the .mdp file settings?? -- Bharat -- gmx-users mailing listgmx-users@gromacs.org http://lists.gromacs.org/mailman/listinfo/gmx-users * Please search the archive at http://www.gromacs.org/Support/Mailing_Lists/Search before posting! * Please don't post (un)subscribe requests to the list. Use the www interface or send it to gmx-users-requ...@gromacs.org. * Can't post? Read http://www.gromacs.org/Support/Mailing_Lists -- gmx-users mailing listgmx-users@gromacs.org http://lists.gromacs.org/mailman/listinfo/gmx-users * Please search the archive at http://www.gromacs.org/Support/Mailing_Lists/Search before posting! * Please don't post (un)subscribe requests to the list. Use the www interface or send it to gmx-users-requ...@gromacs.org. * Can't post? Read http://www.gromacs.org/Support/Mailing_Lists -- gmx-users mailing listgmx-users@gromacs.org http://lists.gromacs.org/mailman/listinfo/gmx-users * Please search the archive at http://www.gromacs.org/Support/Mailing_Lists/Search before posting! * Please don't post (un)subscribe requests to the list. Use the
Re: [Spam:*****] [gmx-users] REMD analysis
You need to increase the temperature gaps indeed if you want acceptance ratio ~0.2/0.3. But again this won't work with the water … It is not clear what happens in your index file but probably a problem from grace to plot so many points … you can try to increase the Max drawing path length in the preference menu of grace. On May 13, 2013, at 4:22 PM, bharat gupta bharat.85.m...@gmail.com wrote: Dear Sir, I repeated the simulation again for 25 replicas with the following temp. distribution . 280 289.1 298.5 308.2 318.2 328.6 339.3 350.3 361.7 373.5 385.6 398.1 411.1 424.4 438.3 452.5 467.2 482.4 498.1 514.3 531.0 548.3 566.1 584.5 603.5 623.2 The output of md.log file is :- Replica exchange statistics Repl 24999 attempts, 12500 odd, 12499 even Repl average probabilities: Repl 0123456789 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 Repl .63 .63 .62 .62 .61 .61 .60 .60 .59 .59 .58 .59 .59 .60 .60 .61 .62 .62 .63 .64 .64 .65 .65 .66 .66 Repl number of exchanges: Repl 0123456789 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 Repl 7822 7752 7816 7760 7639 7628 7511 7442 7375 7332 7312 7424 7408 7410 7522 7559 7684 7697 7878 7927 7917 8073 8151 8208 8266 Repl average number of exchanges: Repl 0123456789 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 Repl .63 .62 .63 .62 .61 .61 .60 .60 .59 .59 .58 .59 .59 .59 .60 .60 .61 .62 .63 .63 .63 .65 .65 .66 .66 The average acceptance ration is around 0.6 which is still high. The link for replica_temp,replica_index : https://www.dropbox.com/s/c7soajnwc3uww8j/replica_temp.png https://www.dropbox.com/s/wvx82m4c6cnsfit/replica_index.png The temp files look better but the index file looks weird ... Do i need to experiment with the gap difference in order to get the required ration of 0.2-0.3 ?? There is some problem with the .mdp file settings?? -- Bharat -- gmx-users mailing listgmx-users@gromacs.org http://lists.gromacs.org/mailman/listinfo/gmx-users * Please search the archive at http://www.gromacs.org/Support/Mailing_Lists/Search before posting! * Please don't post (un)subscribe requests to the list. Use the www interface or send it to gmx-users-requ...@gromacs.org. * Can't post? Read http://www.gromacs.org/Support/Mailing_Lists -- gmx-users mailing listgmx-users@gromacs.org http://lists.gromacs.org/mailman/listinfo/gmx-users * Please search the archive at http://www.gromacs.org/Support/Mailing_Lists/Search before posting! * Please don't post (un)subscribe requests to the list. Use the www interface or send it to gmx-users-requ...@gromacs.org. * Can't post? Read http://www.gromacs.org/Support/Mailing_Lists
Re: [Spam:*****] [gmx-users] REMD analysis
Dear Sir, Here's the result for the REMD trial with large temperature gaps. Temp. distribution : 280.0 294.9 310.7 327.3 344.7 363.1 382.5 402.9 424.4 447.1 471.0 496.1 522.6 550.5 579.9 610.8 Out of md16.log : Replica exchange statistics Repl 249 attempts, 125 odd, 124 even Repl average probabilities: Repl 0123456789 10 11 12 13 14 15 Repl .40 .34 .38 .43 .43 .36 .45 .40 .37 .48 .47 .45 .47 .44 .46 Repl number of exchanges: Repl 0123456789 10 11 12 13 14 15 Repl 50 42 46 52 57 40 58 49 42 53 61 63 56 57 58 Repl average number of exchanges: Repl 0123456789 10 11 12 13 14 15 Repl .40 .34 .37 .42 .46 .32 .46 .40 .34 .43 .49 .51 .45 .46 .46 Average acceptance ratio : 0.46 But, the repli_index.xvg and replica_temp.xvg files still shows that the replicas does not exchange equally well . https://www.dropbox.com/s/zkbwpuj7l2o282b/replica_index.png https://www.dropbox.com/s/0c8gp584v1hvlbx/replica_temp.png what could be wrong in this case?? Is it the mdp file settings or implicit solvent setting. Does the time to replica to exhange also affects their swapping ?? On Tue, May 14, 2013 at 12:24 AM, XAvier Periole x.peri...@rug.nl wrote: You need to increase the temperature gaps indeed if you want acceptance ratio ~0.2/0.3. But again this won't work with the water … It is not clear what happens in your index file but probably a problem from grace to plot so many points … you can try to increase the Max drawing path length in the preference menu of grace. On May 13, 2013, at 4:22 PM, bharat gupta bharat.85.m...@gmail.com wrote: Dear Sir, I repeated the simulation again for 25 replicas with the following temp. distribution . 280 289.1 298.5 308.2 318.2 328.6 339.3 350.3 361.7 373.5 385.6 398.1 411.1 424.4 438.3 452.5 467.2 482.4 498.1 514.3 531.0 548.3 566.1 584.5 603.5 623.2 The output of md.log file is :- Replica exchange statistics Repl 24999 attempts, 12500 odd, 12499 even Repl average probabilities: Repl 0123456789 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 Repl .63 .63 .62 .62 .61 .61 .60 .60 .59 .59 .58 .59 .59 .60 .60 .61 .62 .62 .63 .64 .64 .65 .65 .66 .66 Repl number of exchanges: Repl 0123456789 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 Repl 7822 7752 7816 7760 7639 7628 7511 7442 7375 7332 7312 7424 7408 7410 7522 7559 7684 7697 7878 7927 7917 8073 8151 8208 8266 Repl average number of exchanges: Repl 0123456789 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 Repl .63 .62 .63 .62 .61 .61 .60 .60 .59 .59 .58 .59 .59 .59 .60 .60 .61 .62 .63 .63 .63 .65 .65 .66 .66 The average acceptance ration is around 0.6 which is still high. The link for replica_temp,replica_index : https://www.dropbox.com/s/c7soajnwc3uww8j/replica_temp.png https://www.dropbox.com/s/wvx82m4c6cnsfit/replica_index.png The temp files look better but the index file looks weird ... Do i need to experiment with the gap difference in order to get the required ration of 0.2-0.3 ?? There is some problem with the .mdp file settings?? -- Bharat -- gmx-users mailing listgmx-users@gromacs.org http://lists.gromacs.org/mailman/listinfo/gmx-users * Please search the archive at http://www.gromacs.org/Support/Mailing_Lists/Search before posting! * Please don't post (un)subscribe requests to the list. Use the www interface or send it to gmx-users-requ...@gromacs.org. * Can't post? Read http://www.gromacs.org/Support/Mailing_Lists -- gmx-users mailing listgmx-users@gromacs.org http://lists.gromacs.org/mailman/listinfo/gmx-users * Please search the archive at http://www.gromacs.org/Support/Mailing_Lists/Search before posting! * Please don't post (un)subscribe requests to the list. Use the www interface or send it to gmx-users-requ...@gromacs.org. * Can't post? Read http://www.gromacs.org/Support/Mailing_Lists -- gmx-users mailing listgmx-users@gromacs.org http://lists.gromacs.org/mailman/listinfo/gmx-users * Please search the archive at http://www.gromacs.org/Support/Mailing_Lists/Search before posting! * Please don't post (un)subscribe requests to the list. Use the www interface or send it to gmx-users-requ...@gromacs.org. * Can't post? Read http://www.gromacs.org/Support/Mailing_Lists
