Hi, I am using the optimize_wxc=true and optimize_wxu=true in my last refinement. The R and Rfree went down but the RMSDs were way too large. I wonder if phenix.refine still targets the geometry when it optimizes the R and Rfree. Here are the results from almost two identical runs except the weight optimization. 1) optimize_wxc = False optimize_wxu = False wxc_scale=0.08 wxu_scale=1.66 results: R=0.2294 R-free=0.2709 angl=1.087 bond=0.006 2) optimize_wxc = True optimize_wxu = True wxc_scale=0.08 wxu_scale=1.66 results: R=0.2141 R-free=0.2628 angl=2.894 bond=0.033 I am using the dev-412 version. Thanks. -- Jianghai
Hi Jianghai, try adjusting these values to match your expectations: target_weights { bonds_rmsd_max = 0.05 angles_rmsd_max = 3.5 } I've never really played with this, but this is done exactly to address your question. That weight will be selected that minimizes the Rfree and obeys the above limits. I'm not sure what happens if you already start with the values that are higher than these limits. Please let me know if you have any problems with this. Pavel. On 5/25/10 9:08 AM, Jianghai Zhu wrote:
Hi,
I am using the optimize_wxc=true and optimize_wxu=true in my last refinement. The R and Rfree went down but the RMSDs were way too large. I wonder if phenix.refine still targets the geometry when it optimizes the R and Rfree. Here are the results from almost two identical runs except the weight optimization.
1) optimize_wxc = False optimize_wxu = False wxc_scale=0.08 wxu_scale=1.66
results: R=0.2294 R-free=0.2709 angl=1.087 bond=0.006
2) optimize_wxc = True optimize_wxu = True wxc_scale=0.08 wxu_scale=1.66
results: R=0.2141 R-free=0.2628 angl=2.894 bond=0.033
I am using the dev-412 version. Thanks.
-- Jianghai
_______________________________________________ phenixbb mailing list [email protected] http://phenix-online.org/mailman/listinfo/phenixbb
Hi Pavel -- are those the default targets? Because I've never figured out why phenix.refine consistently comes up with much lower bond-rmsd (0.05-8) than what I thought should be aimed for (0.018ish for high res data -- there are many long ccp4bb threads about this in the archive). phx. On 25/05/2010 23:45, Pavel Afonine wrote:
Hi Jianghai,
try adjusting these values to match your expectations:
target_weights { bonds_rmsd_max = 0.05 angles_rmsd_max = 3.5 }
I've never really played with this, but this is done exactly to address your question. That weight will be selected that minimizes the Rfree and obeys the above limits. I'm not sure what happens if you already start with the values that are higher than these limits.
Please let me know if you have any problems with this.
Pavel.
On 5/25/10 9:08 AM, Jianghai Zhu wrote:
Hi,
I am using the optimize_wxc=true and optimize_wxu=true in my last refinement. The R and Rfree went down but the RMSDs were way too large. I wonder if phenix.refine still targets the geometry when it optimizes the R and Rfree. Here are the results from almost two identical runs except the weight optimization.
1) optimize_wxc = False optimize_wxu = False wxc_scale=0.08 wxu_scale=1.66
results: R=0.2294 R-free=0.2709 angl=1.087 bond=0.006
2) optimize_wxc = True optimize_wxu = True wxc_scale=0.08 wxu_scale=1.66
results: R=0.2141 R-free=0.2628 angl=2.894 bond=0.033
I am using the dev-412 version. Thanks.
-- Jianghai
_______________________________________________ phenixbb mailing list [email protected] http://phenix-online.org/mailman/listinfo/phenixbb
phenixbb mailing list [email protected] http://phenix-online.org/mailman/listinfo/phenixbb
Yes, these limits fixed the problem. But the default values should be 0.02 and 2, respectively, I believe. -- Jianghai On May 25, 2010, at 6:45 PM, Pavel Afonine wrote:
Hi Jianghai,
try adjusting these values to match your expectations:
target_weights { bonds_rmsd_max = 0.05 angles_rmsd_max = 3.5 }
I've never really played with this, but this is done exactly to address your question. That weight will be selected that minimizes the Rfree and obeys the above limits. I'm not sure what happens if you already start with the values that are higher than these limits.
Please let me know if you have any problems with this.
Pavel.
On 5/25/10 9:08 AM, Jianghai Zhu wrote:
Hi,
I am using the optimize_wxc=true and optimize_wxu=true in my last refinement. The R and Rfree went down but the RMSDs were way too large. I wonder if phenix.refine still targets the geometry when it optimizes the R and Rfree. Here are the results from almost two identical runs except the weight optimization.
1) optimize_wxc = False optimize_wxu = False wxc_scale=0.08 wxu_scale=1.66
results: R=0.2294 R-free=0.2709 angl=1.087 bond=0.006
2) optimize_wxc = True optimize_wxu = True wxc_scale=0.08 wxu_scale=1.66
results: R=0.2141 R-free=0.2628 angl=2.894 bond=0.033
I am using the dev-412 version. Thanks.
-- Jianghai
_______________________________________________ phenixbb mailing list [email protected] http://phenix-online.org/mailman/listinfo/phenixbb
_______________________________________________ phenixbb mailing list [email protected] http://phenix-online.org/mailman/listinfo/phenixbb
Hi Jianghai,
Yes, these limits fixed the problem.
thanks for letting know, I'm glad it worked for you!
But the default values should be 0.02 and 2, respectively, I believe.
Well, I guess better to have it resolution dependent, and best to have it local per bond or so. I have no intention to open another can of worms, but where those 0.02 and 2 come from? Can you really see that level of details at say 2.5A resolution and lower? I guess at that resolutions they should be both zero. In contrast, at higher resolution larger deviations are justified just by the data and so larger margins in restraints should be allowed. I know there is a heap of good papers on this matter, including the recent ones, and there are long threads on bb, but to me this question is still somewhat open. May be someone on the board can thoroughly/convincingly comment on this, summarizing all previous discussions? Anyway, I agree, 0.05 and 3.5 is probably too generous. All the best! Pavel.
On Wed, May 26, 2010 at 8:30 AM, Pavel Afonine
Well, I guess better to have it resolution dependent, and best to have it local per bond or so. I have no intention to open another can of worms, but where those 0.02 and 2 come from? Can you really see that level of details at say 2.5A resolution and lower? I guess at that resolutions they should be both zero. In contrast, at higher resolution larger deviations are justified just by the data and so larger margins in restraints should be allowed. I know there is a heap of good papers on this matter, including the recent ones, and there are long threads on bb, but to me this question is still somewhat open. May be someone on the board can thoroughly/convincingly comment on this, summarizing all previous discussions?
My guess would be that the limits are derived from Engh & Huber. I was taught to use 0.016 and 1.6; 0.02 and 2 are a bit on the high end but for very high-resolution structures this wouldn't be inappropriate, and it's arguably helpful earlier in refinement to let the geometry be a bit looser. I don't think it's practical to force them to be zero at high resolution - it will make it much harder to converge during refinement. Regardless of what the limits *should* be at different resolutions, reviewers will expect something below 0.02/2, and the overwhelming majority of protein structures will be done at resolutions where these values are indeed appropriate at upper limits. -Nat
Are these limits the max values for a single bond or angle? or they are the targets for the whole structure? If they are the latter, then 0.02 and 2 came from the classic Engh and Huber paper. And I agree that the lower the resolution, the tighten these constraints should be. -- Jianghai On May 26, 2010, at 11:30 AM, Pavel Afonine wrote:
Hi Jianghai,
Yes, these limits fixed the problem.
thanks for letting know, I'm glad it worked for you!
But the default values should be 0.02 and 2, respectively, I believe.
Well, I guess better to have it resolution dependent, and best to have it local per bond or so. I have no intention to open another can of worms, but where those 0.02 and 2 come from? Can you really see that level of details at say 2.5A resolution and lower? I guess at that resolutions they should be both zero. In contrast, at higher resolution larger deviations are justified just by the data and so larger margins in restraints should be allowed. I know there is a heap of good papers on this matter, including the recent ones, and there are long threads on bb, but to me this question is still somewhat open. May be someone on the board can thoroughly/convincingly comment on this, summarizing all previous discussions?
Anyway, I agree, 0.05 and 3.5 is probably too generous.
All the best! Pavel.
_______________________________________________ phenixbb mailing list [email protected] http://phenix-online.org/mailman/listinfo/phenixbb
After a few trials, looks like the angles_rmsd_max cannot be set to a value lower than 1.0. Everytime I did that, the values for bonds_rmsd_max and angles_rmsd_max reverted to the default numbers in the refinement and the limits I set in the input file are not effective. How do I fix that? Thanks. -- Jianghai On May 26, 2010, at 11:30 AM, Pavel Afonine wrote:
Hi Jianghai,
Yes, these limits fixed the problem.
thanks for letting know, I'm glad it worked for you!
But the default values should be 0.02 and 2, respectively, I believe.
Well, I guess better to have it resolution dependent, and best to have it local per bond or so. I have no intention to open another can of worms, but where those 0.02 and 2 come from? Can you really see that level of details at say 2.5A resolution and lower? I guess at that resolutions they should be both zero. In contrast, at higher resolution larger deviations are justified just by the data and so larger margins in restraints should be allowed. I know there is a heap of good papers on this matter, including the recent ones, and there are long threads on bb, but to me this question is still somewhat open. May be someone on the board can thoroughly/convincingly comment on this, summarizing all previous discussions?
Anyway, I agree, 0.05 and 3.5 is probably too generous.
All the best! Pavel.
_______________________________________________ phenixbb mailing list [email protected] http://phenix-online.org/mailman/listinfo/phenixbb
Hi Jianghai, I have no quick answer. I will have a look within next few days and let you know. It may be a side effect of runtime optimization - I'm trying to save some refinement runs by looking at the available accumulated statistics and extrapolating from that. Anyway, I guess it's time to review that code (given the amount of usage). May be it is safer to give up some runtime over the robustness and predictability... will see. Have a great week-end! Pavel. On 5/28/10 9:41 PM, Jianghai Zhu wrote:
After a few trials, looks like the angles_rmsd_max cannot be set to a value lower than 1.0. Everytime I did that, the values for bonds_rmsd_max and angles_rmsd_max reverted to the default numbers in the refinement and the limits I set in the input file are not effective. How do I fix that? Thanks.
-- Jianghai
On May 26, 2010, at 11:30 AM, Pavel Afonine wrote:
Hi Jianghai,
Yes, these limits fixed the problem.
thanks for letting know, I'm glad it worked for you!
But the default values should be 0.02 and 2, respectively, I believe.
Well, I guess better to have it resolution dependent, and best to have it local per bond or so. I have no intention to open another can of worms, but where those 0.02 and 2 come from? Can you really see that level of details at say 2.5A resolution and lower? I guess at that resolutions they should be both zero. In contrast, at higher resolution larger deviations are justified just by the data and so larger margins in restraints should be allowed. I know there is a heap of good papers on this matter, including the recent ones, and there are long threads on bb, but to me this question is still somewhat open. May be someone on the board can thoroughly/convincingly comment on this, summarizing all previous discussions?
Anyway, I agree, 0.05 and 3.5 is probably too generous.
All the best! Pavel.
_______________________________________________ phenixbb mailing list [email protected] http://phenix-online.org/mailman/listinfo/phenixbb
_______________________________________________ phenixbb mailing list [email protected] http://phenix-online.org/mailman/listinfo/phenixbb
participants (4)
-
Frank von Delft -
Jianghai Zhu -
Nathaniel Echols -
Pavel Afonine