Re: [ccp4bb] Refmac5 refinement question
Hi Ray, Not necessarily close to 1.00 but rather 1.00 or lower. Cheers, Robbie -Original Message- From: Raymond Brown [mailto:ray-br...@att.net] Sent: Tuesday, March 12, 2019 15:56 To: Robbie Joosten Cc: CCP4BB@JISCMAIL.AC.UK Subject: Re: [ccp4bb] Refmac5 refinement question Hi Robbie, Thank you for your very clear explanation and suggestions. Am I right in thinking that what this means is that the zBOND and zANGLE values in Refmac5 refinement should be made to be close to 1.000? Best Ray On Tue, 3/12/19, Robbie Joosten wrote: Subject: Re: [ccp4bb] Refmac5 refinement question To: CCP4BB@JISCMAIL.AC.UK Date: Tuesday, March 12, 2019, 4:58 AM Hi Ray, This is how I see it: Because different bond length and angle target tolerances/sigmas you cannot compare them on an absolute scale. What is less likely? A two 0.020A deviation in the CD1-CG in PHE or a the same deviation in the CD1-CG bond in LEU. If you think they are equally likely, you can use rmsd, if not you probably realise that you are comparing apples and oranges. Enter rmsZ. Instead of comparing the absolute deviations you compare the deviations expressed in sigma (the Z-score). This brings everything on the same scale so you can actually compare them (provided the restraints themselves, especially the sigmas, make sense). Now let's assume that your deviations (expressed as Z-scores) have a normal distribution. Then the rmsZ says something about the broadness of this distribution. With rmsZ equal to 1 your distribution is as wide as that of the source of the restraints (i.e. small molecule structures in the COD, in the case of Refmac). Assuming that your data are a bit worse, you can argue that you do not have any strong evidence to allow wider distributions. That means your rmsZ should not be greater than 1. There seems to be a bit of a downward trend with resolution and rmsZ: lower resolution, lower rmsZ. However the correlation is not very strong and therefor I would say that any rmsZ < 1.00 is fine. Refmac gives these numbers for bond lengths and angles, not for chiral volumes, but those all have the same sigma (AFAIK). Don't go for hard numbers with deviations, unless you find a good rationale to do so (if you do, please share it). I've seen too many case of grossly over/underrestrained models where a much better fit to the data (and other validation criteria) can be obtained by just playing a bit with the geometric restraint weights. HTH, Robbie -Original Message- From: CCP4 bulletin board [mailto:CCP4BB@JISCMAIL.AC.UK] On Behalf Of Raymond Brown Sent: Monday, March 11, 2019 17:11 To: CCP4BB@JISCMAIL.AC.UK Subject: [ccp4bb] Refmac5 refinement question Hi folks, What are acceptable values for RMS Bond length, RMS Bond angle and RMS Chiral volume? The tutorial suggests RMS Bond length of 0.0200? I would like to hear your suggestions and/or rationale. Many thanks Ray Brown To unsubscribe from the CCP4BB list, click the following link: https://www.jiscmail.ac.uk/cgi-bin/webadmin?SUBED1=CCP4BB=1 To unsubscribe from the CCP4BB list, click the following link: https://www.jiscmail.ac.uk/cgi-bin/webadmin?SUBED1=CCP4BB=1 To unsubscribe from the CCP4BB list, click the following link: https://www.jiscmail.ac.uk/cgi-bin/webadmin?SUBED1=CCP4BB=1
Re: [ccp4bb] Refmac5 refinement question
Hi Robbie, Thank you for your very clear explanation and suggestions. Am I right in thinking that what this means is that the zBOND and zANGLE values in Refmac5 refinement should be made to be close to 1.000? Best Ray On Tue, 3/12/19, Robbie Joosten wrote: Subject: Re: [ccp4bb] Refmac5 refinement question To: CCP4BB@JISCMAIL.AC.UK Date: Tuesday, March 12, 2019, 4:58 AM Hi Ray, This is how I see it: Because different bond length and angle target tolerances/sigmas you cannot compare them on an absolute scale. What is less likely? A two 0.020A deviation in the CD1-CG in PHE or a the same deviation in the CD1-CG bond in LEU. If you think they are equally likely, you can use rmsd, if not you probably realise that you are comparing apples and oranges. Enter rmsZ. Instead of comparing the absolute deviations you compare the deviations expressed in sigma (the Z-score). This brings everything on the same scale so you can actually compare them (provided the restraints themselves, especially the sigmas, make sense). Now let's assume that your deviations (expressed as Z-scores) have a normal distribution. Then the rmsZ says something about the broadness of this distribution. With rmsZ equal to 1 your distribution is as wide as that of the source of the restraints (i.e. small molecule structures in the COD, in the case of Refmac). Assuming that your data are a bit worse, you can argue that you do not have any strong evidence to allow wider distributions. That means your rmsZ should not be greater than 1. There seems to be a bit of a downward trend with resolution and rmsZ: lower resolution, lower rmsZ. However the correlation is not very strong and therefor I would say that any rmsZ < 1.00 is fine. Refmac gives these numbers for bond lengths and angles, not for chiral volumes, but those all have the same sigma (AFAIK). Don't go for hard numbers with deviations, unless you find a good rationale to do so (if you do, please share it). I've seen too many case of grossly over/underrestrained models where a much better fit to the data (and other validation criteria) can be obtained by just playing a bit with the geometric restraint weights. HTH, Robbie -Original Message- From: CCP4 bulletin board [mailto:CCP4BB@JISCMAIL.AC.UK] On Behalf Of Raymond Brown Sent: Monday, March 11, 2019 17:11 To: CCP4BB@JISCMAIL.AC.UK Subject: [ccp4bb] Refmac5 refinement question Hi folks, What are acceptable values for RMS Bond length, RMS Bond angle and RMS Chiral volume? The tutorial suggests RMS Bond length of 0.0200? I would like to hear your suggestions and/or rationale. Many thanks Ray Brown To unsubscribe from the CCP4BB list, click the following link: https://www.jiscmail.ac.uk/cgi-bin/webadmin?SUBED1=CCP4BB=1 To unsubscribe from the CCP4BB list, click the following link: https://www.jiscmail.ac.uk/cgi-bin/webadmin?SUBED1=CCP4BB=1 To unsubscribe from the CCP4BB list, click the following link: https://www.jiscmail.ac.uk/cgi-bin/webadmin?SUBED1=CCP4BB=1
Re: [ccp4bb] Refmac5 refinement question
Hi Ray, This is how I see it: Because different bond length and angle target tolerances/sigmas you cannot compare them on an absolute scale. What is less likely? A two 0.020A deviation in the CD1-CG in PHE or a the same deviation in the CD1-CG bond in LEU. If you think they are equally likely, you can use rmsd, if not you probably realise that you are comparing apples and oranges. Enter rmsZ. Instead of comparing the absolute deviations you compare the deviations expressed in sigma (the Z-score). This brings everything on the same scale so you can actually compare them (provided the restraints themselves, especially the sigmas, make sense). Now let's assume that your deviations (expressed as Z-scores) have a normal distribution. Then the rmsZ says something about the broadness of this distribution. With rmsZ equal to 1 your distribution is as wide as that of the source of the restraints (i.e. small molecule structures in the COD, in the case of Refmac). Assuming that your data are a bit worse, you can argue that you do not have any strong evidence to allow wider distributions. That means your rmsZ should not be greater than 1. There seems to be a bit of a downward trend with resolution and rmsZ: lower resolution, lower rmsZ. However the correlation is not very strong and therefor I would say that any rmsZ < 1.00 is fine. Refmac gives these numbers for bond lengths and angles, not for chiral volumes, but those all have the same sigma (AFAIK). Don't go for hard numbers with deviations, unless you find a good rationale to do so (if you do, please share it). I've seen too many case of grossly over/underrestrained models where a much better fit to the data (and other validation criteria) can be obtained by just playing a bit with the geometric restraint weights. HTH, Robbie -Original Message- From: CCP4 bulletin board [mailto:CCP4BB@JISCMAIL.AC.UK] On Behalf Of Raymond Brown Sent: Monday, March 11, 2019 17:11 To: CCP4BB@JISCMAIL.AC.UK Subject: [ccp4bb] Refmac5 refinement question Hi folks, What are acceptable values for RMS Bond length, RMS Bond angle and RMS Chiral volume? The tutorial suggests RMS Bond length of 0.0200? I would like to hear your suggestions and/or rationale. Many thanks Ray Brown To unsubscribe from the CCP4BB list, click the following link: https://www.jiscmail.ac.uk/cgi-bin/webadmin?SUBED1=CCP4BB=1 To unsubscribe from the CCP4BB list, click the following link: https://www.jiscmail.ac.uk/cgi-bin/webadmin?SUBED1=CCP4BB=1
[ccp4bb] Refmac5 refinement question
Hi folks, What are acceptable values for RMS Bond length, RMS Bond angle and RMS Chiral volume? The tutorial suggests RMS Bond length of 0.0200? I would like to hear your suggestions and/or rationale. Many thanks Ray Brown To unsubscribe from the CCP4BB list, click the following link: https://www.jiscmail.ac.uk/cgi-bin/webadmin?SUBED1=CCP4BB=1
