Hi Vipul and Qiaoling, Setting occupancies to 0 will give somewhat misleading results. It is good to test whether density will disappear or not, but not for the final model. Users of the model might misinterpret it. Especially if you start leaving out 'random' poorly fitting atoms.
High B-factors can occur and the average seems fine. High variance can be an indication of too loose B-factor restraints. Since you asked for suggestions of other protocols on the CCP4bb: Have you tried Refmac? And a shameless plug: Have you tried pdb-redo? Among other things, this tries to find a good protocol to get the best performance out of Refmac. Cheers, Robbie > -----Original Message----- > From: CCP4 bulletin board [mailto:CCP4BB@JISCMAIL.AC.UK] On Behalf Of > Vipul Panchal > Sent: Thursday, May 04, 2017 13:45 > To: CCP4BB@JISCMAIL.AC.UK > Subject: Re: [ccp4bb] very high B-factor > > 1) the difference is 5.23 % and it is good and acceptable. > 2) if higher value belongs to residue, then you need to change the rotamer > Using coot , follow menu-->distance-->residue info. You will get to know to > whom 417 B factor value belongs. Meanwhile, just check if modeled residue > or part of residue has relevant density. If density is missing for part of > residue, then set occupancy to zero for the relevant atom(s). If density is > missing for the model of residue, then just remove side chain. I would > recommend to check density fit (coot-->menu--> validate--> density fit) for > all residues with >50 B factor value (coot-->menu--> validate--> temp. fact. > variance analysis). > > > On 04-May-2017 3:26 PM, "yanqiaoling2782048" > <yanqiaoling2782...@126.com> wrote: > > > Hi Vipul, > > Thanks for your quick reply. > > 1. Actually, i mainly comparing the gap between Rwork and Rfree > which should be <5%, and then the overall b-factor, RMSD of bond and > angle. For me, no5 is the best one, but the "B factor variance Graphs" make > me uncomfortable. The R-factor gap of no6 is 6.23%, does this mean > overfitting? > > 2. The b-factor value of 76 is belong to the atom of the residue > obtained from the pdb file. I'm not sure whether the B-factor in "B factor > variance Graphs" belong to residue or atom, but i guess 417 belongs to > residue. What's the relationship of the two factor? > > > Best regards, > Qiaoling Yan > > > > > At 2017-05-04 15:58:36, "Vipul Panchal" <panchal.vi...@igib.in> > wrote: > > > Hi, > > 1> > Well, it seems you are just comparing Max value across all > protocols. You should compare average values as it also takes into account > no. of atoms with given values of B factor. As per refinement results, it > seems though no. 6 has highest B factor value, number of atomic outliers are > relatively low. So if you ask me no. 6 is the best one. > Coot results are also in congruent with refinement results. I > think you should go ahead with no.6 is strategy. > > 2> > When comparing B factor values of 417 vs 76, you should > consider what it belongs to. As i may guess, 417 value belongs to an atom of > residue whereas 76 belongs to residue. Meaning 417 is individual B factor > whereas 76 is grouped B factor. > > > On Thu, May 4, 2017 at 9:08 AM, yanqiaoling2782048 > <yanqiaoling2782...@126.com> wrote: > > > Dear all, > > I'm working on a crystal structure with resolution of > 2.2A. At the final step, I use different strategies to refine the structure, > they > are: > no4: strategy=individual_sites+individual_adp+tls / > set_b_iso=20 > results: Rwork/free=0.2052/0.2658 b- > factor=11.4/136.8/48(min/max/average) > > no5: strategy=individual_sites+individual_adp+tls / > set_b_iso=10 / optimize_xyz/adp_weight=true > results: Rwork/free=0.2161/0.2639 b- > factor=11.3/135.2/48.4(min/max/average) > > no6: strategy=individual_sites+individual_adp / > anisotropic for all residues and isotropic for water / > set_b_iso=10 / optimize_xyz/adp_weight=true > results: Rwork/free=0.2183/0.2706 b- > factor=10.9/144.6/45.7 (min/max/average) > PS: the results is read from pdb file > > The results showed that the strategy of no5 is the > best one. But the "B factor variance Graphs" generated by coot with > menu/validate/temp.fact.variance analysis, have shown that no6 have the > lowest B-factors (the attached figure is the B-factor graphs of three pdb > files). And my questions are: > 1. Which strategy should I choose to refine my > structure? Or any other suggestions to refine the structure at 2.2A > resolution? > 2. Does it possible that some residues have very high > B-factor in "B factor variance Graphs", while in the pdb fileļ¼ the b-factor of > corresponding residues are relatively low? For example, one residue have B- > factors of 417 in "B factor variance Graphs", but in PDB file the b factor is > 76. > Does the two factor mean the same thing? > 3. If i want to set the weight manually, which > parameter should i set, wxc/wxc_scale? or others? > > Thanks in advance. > > Best regards, > Qiaoling Yan > > > > > > > > > > -- > > Vipul Panchal > Senior Research Fellow, > Respiratory disease and biology, > CSIR-IGIB > (M)-9540113372 > > > > > >
