Re: [phenixbb] Rfree increases during ordered solvent filtering
Hi Pavel, Thanks. I don't believe that is the cause of this refinement anomaly. If I run the same job without the filter_only option, I am still seeing a very large jump in R/Rfree at the ordered solvent location/analysis/update step (before 2_xyz). ordered_solvent=true 0 : 0.1564 0.1709 2.905699e+00 3.076326e+00 1_bss: 0.1145 0.1431 2.837441e+00 3.027552e+00 1_xyz: 0.1159 0.1433 2.846219e+00 3.029651e+00 1_adp: 0.1159 0.1433 2.846219e+00 3.029651e+00 1_occ: 0.1155 0.1429 2.844726e+00 3.028603e+00 2_bss: 0.1154 0.1427 2.844702e+00 3.028393e+00 2_xyz: 0.1313 0.1555 2.950732e+00 3.111378e+00 2_adp: 0.1307 0.1543 2.945031e+00 3.101981e+00 2_occ: 0.1306 0.1542 2.945015e+00 3.101954e+00 3_bss: 0.1301 0.1538 2.944930e+00 3.101580e+00 3_xyz: 0.1312 0.1565 2.951306e+00 3.117549e+00 3_adp: 0.1306 0.1546 2.940355e+00 3.100448e+00 3_occ: 0.1305 0.1546 2.940347e+00 3.100439e+00 3_bss: 0.1292 0.1534 2.940856e+00 3.100060e+00 However, if I run this job with no ordered solvent manipulation, the R/Rfree are stable at their previous lower value. ordered_solvent=false 0 : 0.1564 0.1709 2.905699e+00 3.076326e+00 1_bss: 0.1145 0.1431 2.837441e+00 3.027552e+00 1_xyz: 0.1159 0.1433 2.846219e+00 3.029651e+00 1_adp: 0.1159 0.1433 2.846219e+00 3.029651e+00 1_occ: 0.1155 0.1429 2.844726e+00 3.028603e+00 2_bss: 0.1154 0.1427 2.844702e+00 3.028393e+00 2_xyz: 0.1151 0.1427 2.842538e+00 3.027628e+00 2_adp: 0.1151 0.1427 2.842538e+00 3.027628e+00 2_occ: 0.1151 0.1427 2.842433e+00 3.027654e+00 3_bss: 0.1151 0.1426 2.842430e+00 3.027578e+00 3_xyz: 0.1156 0.1428 2.846107e+00 3.029162e+00 3_adp: 0.1156 0.1428 2.846106e+00 3.029160e+00 3_occ: 0.1156 0.1429 2.846122e+00 3.029225e+00 3_bss: 0.1156 0.1429 2.846122e+00 3.029225e+00 I wouldn't expect to see a 1% jump in Rfree after minor tweaking of a few dozen waters. This is not a small structure. There must be something strange in the refinement protocol during the ordered water treatment. I had thought it could be a weighting issue, but subsequent refinement with optimization of wxc and wxu does not bring the R/Rfree down to the original range. Thanks, Jennifer (Running version 2008_05_03_2330) ................................................... Pavel Afonine writes:
Hi Jennifer,
you are using "ordered_solvent.mode=filter_only". This means that the new waters are not added but the existing ones are filtered (removed) at final step based on selection criteria (B-factors, distances, H-bonding, etc...). Do you really want this? If not, then do not use "ordered_solvent.mode=filter_only" (the default is "auto" which is good in most of cases).
Please let me know if you have any questions!
Pavel.
On 6/24/2008 2:42 PM, Jennifer L Ekstrom wrote:
I am running phenix.refine (version 2008_05_03_2330) with the command-line options ordered_solvent=true ordered_solvent.mode=filter_only
During the final step (3_bss), my R and Rfree increase significantly. Any thoughts on why this jump would occcur? Per the notations in the logfile, only 30 waters were removed in the final model.
Thanks, Jennifer
R-factors, x-ray target values and norm of gradient of x-ray target stage r-work r-free xray_target_w xray_target_t 0 : 0.1564 0.1709 2.905699e+00 3.076326e+00 1_bss: 0.1145 0.1431 2.837441e+00 3.027552e+00 1_xyz: 0.1159 0.1433 2.846219e+00 3.029651e+00 1_adp: 0.1159 0.1433 2.846219e+00 3.029651e+00 1_occ: 0.1155 0.1429 2.844726e+00 3.028603e+00 2_bss: 0.1154 0.1427 2.844702e+00 3.028393e+00 2_xyz: 0.1151 0.1427 2.842538e+00 3.027628e+00 2_adp: 0.1151 0.1427 2.842538e+00 3.027628e+00 2_occ: 0.1151 0.1427 2.842433e+00 3.027654e+00 3_bss: 0.1151 0.1426 2.842430e+00 3.027578e+00 3_xyz: 0.1156 0.1428 2.846107e+00 3.029162e+00 3_adp: 0.1156 0.1428 2.846106e+00 3.029160e+00 3_occ: 0.1156 0.1429 2.846122e+00 3.029225e+00 3_bss: 0.1306 0.1512 2.943092e+00 3.079441e+00
_______________________________________________ phenixbb mailing list [email protected] http://www.phenix-online.org/mailman/listinfo/phenixbb
Hi Jennifer, yes, I see it now: there must be something weird indeed. At this point I can't tell you much without playing with the refinement myself. Is it possible that you send me the data and model (to my address, not to the whole phenixbb) so I can reproduce this behavior and possibly fix it or suggest how to proceed? I promise to keep the files strictly confidential and remove them from my computer once the issue is fixed. Pavel. On 6/25/2008 1:05 PM, Jennifer L Ekstrom wrote:
Hi Pavel, Thanks. I don't believe that is the cause of this refinement anomaly. If I run the same job without the filter_only option, I am still seeing a very large jump in R/Rfree at the ordered solvent location/analysis/update step (before 2_xyz). ordered_solvent=true 0 : 0.1564 0.1709 2.905699e+00 3.076326e+00 1_bss: 0.1145 0.1431 2.837441e+00 3.027552e+00 1_xyz: 0.1159 0.1433 2.846219e+00 3.029651e+00 1_adp: 0.1159 0.1433 2.846219e+00 3.029651e+00 1_occ: 0.1155 0.1429 2.844726e+00 3.028603e+00 2_bss: 0.1154 0.1427 2.844702e+00 3.028393e+00 2_xyz: 0.1313 0.1555 2.950732e+00 3.111378e+00 2_adp: 0.1307 0.1543 2.945031e+00 3.101981e+00 2_occ: 0.1306 0.1542 2.945015e+00 3.101954e+00 3_bss: 0.1301 0.1538 2.944930e+00 3.101580e+00 3_xyz: 0.1312 0.1565 2.951306e+00 3.117549e+00 3_adp: 0.1306 0.1546 2.940355e+00 3.100448e+00 3_occ: 0.1305 0.1546 2.940347e+00 3.100439e+00 3_bss: 0.1292 0.1534 2.940856e+00 3.100060e+00 However, if I run this job with no ordered solvent manipulation, the R/Rfree are stable at their previous lower value. ordered_solvent=false 0 : 0.1564 0.1709 2.905699e+00 3.076326e+00 1_bss: 0.1145 0.1431 2.837441e+00 3.027552e+00 1_xyz: 0.1159 0.1433 2.846219e+00 3.029651e+00 1_adp: 0.1159 0.1433 2.846219e+00 3.029651e+00 1_occ: 0.1155 0.1429 2.844726e+00 3.028603e+00 2_bss: 0.1154 0.1427 2.844702e+00 3.028393e+00 2_xyz: 0.1151 0.1427 2.842538e+00 3.027628e+00 2_adp: 0.1151 0.1427 2.842538e+00 3.027628e+00 2_occ: 0.1151 0.1427 2.842433e+00 3.027654e+00 3_bss: 0.1151 0.1426 2.842430e+00 3.027578e+00 3_xyz: 0.1156 0.1428 2.846107e+00 3.029162e+00 3_adp: 0.1156 0.1428 2.846106e+00 3.029160e+00 3_occ: 0.1156 0.1429 2.846122e+00 3.029225e+00 3_bss: 0.1156 0.1429 2.846122e+00 3.029225e+00
I wouldn't expect to see a 1% jump in Rfree after minor tweaking of a few dozen waters. This is not a small structure. There must be something strange in the refinement protocol during the ordered water treatment. I had thought it could be a weighting issue, but subsequent refinement with optimization of wxc and wxu does not bring the R/Rfree down to the original range. Thanks, Jennifer (Running version 2008_05_03_2330) ...................................................
Pavel Afonine writes:
Hi Jennifer, you are using "ordered_solvent.mode=filter_only". This means that the new waters are not added but the existing ones are filtered (removed) at final step based on selection criteria (B-factors, distances, H-bonding, etc...). Do you really want this? If not, then do not use "ordered_solvent.mode=filter_only" (the default is "auto" which is good in most of cases). Please let me know if you have any questions! Pavel.
On 6/24/2008 2:42 PM, Jennifer L Ekstrom wrote:
I am running phenix.refine (version 2008_05_03_2330) with the command-line options ordered_solvent=true ordered_solvent.mode=filter_only During the final step (3_bss), my R and Rfree increase significantly. Any thoughts on why this jump would occcur? Per the notations in the logfile, only 30 waters were removed in the final model. Thanks, Jennifer
R-factors, x-ray target values and norm of gradient of x-ray target stage r-work r-free xray_target_w xray_target_t 0 : 0.1564 0.1709 2.905699e+00 3.076326e+00 1_bss: 0.1145 0.1431 2.837441e+00 3.027552e+00 1_xyz: 0.1159 0.1433 2.846219e+00 3.029651e+00 1_adp: 0.1159 0.1433 2.846219e+00 3.029651e+00 1_occ: 0.1155 0.1429 2.844726e+00 3.028603e+00 2_bss: 0.1154 0.1427 2.844702e+00 3.028393e+00 2_xyz: 0.1151 0.1427 2.842538e+00 3.027628e+00 2_adp: 0.1151 0.1427 2.842538e+00 3.027628e+00 2_occ: 0.1151 0.1427 2.842433e+00 3.027654e+00 3_bss: 0.1151 0.1426 2.842430e+00 3.027578e+00 3_xyz: 0.1156 0.1428 2.846107e+00 3.029162e+00 3_adp: 0.1156 0.1428 2.846106e+00 3.029160e+00 3_occ: 0.1156 0.1429 2.846122e+00 3.029225e+00 3_bss: 0.1306 0.1512 2.943092e+00 3.079441e+00
_______________________________________________ phenixbb mailing list [email protected] http://www.phenix-online.org/mailman/listinfo/phenixbb
participants (2)
-
Jennifer L Ekstrom -
Pavel Afonine