Re: [ccp4bb] Low resolution and high anisotropy
The improvement in statistics by STARANISO is really spectacular, but I do wonder how much is caused by the directionally variable resolution limits and how much by the removal of solvent rings? The original data set statistics have big dips in CC1/2 at 6, 4.3 and 3.7 Å. Also in the final statistics table, the completeness is really high, but I guess reflections from the "bad" direction(s) are not taken into account? It would be good to also know the spherical completeness and the CC1/2 values before going too far ahead with these data. Or perhaps better phrased: I'd simultaneously try to get a dataset from a better flash-cooled crystal while trying to get the most out of this dataset. Mark J van Raaij Dpto de Estructura de Macromoleculas, lab 20B Centro Nacional de Biotecnologia - CSIC calle Darwin 3 E-28049 Madrid, Spain tel. +34 91 585 4616 (internal 432092) Section Editor Acta Crystallographica F https://journals.iucr.org/f/ > On 24 Oct 2022, at 03:43, Xu, Shenyuan wrote: > > Dear CCP4 community, > > I have encountered a dataset, which I thought should be easy to solve. The > volume of the cell unit seems to be expanded after image 271, which I think > is caused by radiation damage. After removing the last few images, the scaled > statistics seem good with the resolution set at 3:07 A: > > d_max d_min #obs #uniq mult. %comp r_mrg r_meas >r_pim r_anom cc1/2 cc_ano > 91.54 8.33 3086 11972.58 98.44 345.251.30.101 > 0.1330.0860.145 0.973* -0.148 > 8.33 6.61 3005 12252.45 99.51 256.527.30.160 > 0.2120.1380.260 0.923* -0.140 > 6.61 5.78 2883 11902.42 98.43 92.9 9.40.278 > 0.3680.2370.461 0.541* -0.165 > 5.78 5.25 3195 12172.63 99.10 75.6 8.10.283 > 0.3690.2330.436 0.618* -0.107 > 5.25 4.87 3192 12062.65 99.42 92.5 8.30.281 > 0.3670.2320.427 0.856* -0.187 > 4.87 4.59 3230 12132.66 99.02 118.410.30.288 > 0.3780.2400.456 0.870* -0.053 > 4.59 4.36 2797 11562.42 92.93 137.012.20.309 > 0.4100.2660.496 0.843* -0.239 > 4.36 4.17 2694 11182.41 93.17 280.216.20.419 > 0.5750.3920.847 0.491* -0.142 > 4.17 4.01 3102 11882.61 95.50 133.6 9.90.392 > 0.5160.3300.622 0.663* -0.268 > 4.01 3.87 3224 12052.68 98.69 120.4 7.30.406 > 0.5390.3490.687 0.791* -0.126 > 3.87 3.75 2939 11812.49 98.91 90.7 6.80.491 > 0.6520.4250.810 0.691* -0.042 > 3.75 3.64 1981 10211.94 82.34 87.3 5.00.557 > 0.7560.5060.885 0.540* -0.018 > 3.64 3.54 2374 10822.19 89.13 194.611.20.502 > 0.6750.4470.973 0.625* -0.237 > 3.54 3.46 2622 11222.34 92.35 310.510.10.432 > 0.5850.3920.851 0.603* -0.154 > 3.46 3.38 17399451.84 76.58 48.2 3.60.930 > 1.2470.8221.542 0.444* -0.046 > 3.38 3.31 2847 12402.30 98.57 85.4 4.20.691 > 0.9300.6161.242 0.523* -0.060 > 3.31 3.24 2838 11532.46 97.88 70.8 3.30.693 > 0.9340.6201.341 0.362* 0.006 > 3.24 3.18 3097 12122.56 97.66 71.4 2.50.692 > 0.9240.6051.173 0.526* 0.044 > 3.18 3.12 3204 12162.63 99.10 79.6 3.70.668 > 0.8860.5761.338 0.440* -0.081 > 3.12 3.07 3059 11722.61 98.16 73.5 2.40.714 > 0.9450.6121.633 0.362* -0.177 > 91.50 3.07 57108 232592.46 95.22 138.410.80.383 > 0.5130.3360.671 0.656* -0.149 > > I used Mrbump to do the MR, most sequence identities of the starting > templates are more than 0.85, and some of them are structures predicted from > alpha fold 2. But after refinement (including jelly-body, proSmart, TLC), the > best R/Free R stuck at around 0.42/0.45. Inspecting the electron density map > shows that the model does not fit the electron density well. The space group > is P1, Cell is 58.49 63.97 91.60 91.71 91.86 99.47, should be 5 or 6 > molecules in the asymmetric unit. > > I checked the data quality, it said the data is highly anisotropy. Then > I searched the CCP4 forum and used the STARANISO Server and UCLA server, but > still cannot improve the refinement. The data statistics after drawing > ellipsoidal resolution limits is good: > > > resolution observed redundancycompletenesrmerge > i/sigma > before/after before/after before/afterbefore/after > before/after > 7.90 4957 1185 3.5 3.397.3% 92.6%13.0% 12.7%11.9 > 7.2 > 5.5
Re: [ccp4bb] Low resolution and high anisotropy
Dear Shenyuan Xu, it is a good sign that the model and density to not match - at 3.07A, one often only has model bias with a poor MR solution. You don't want to start with refinement, better to start with model building: make the model fit the density better, before you run refinement, and check again for R/Rfree. Best regards, Tim On Mon, 24 Oct 2022 09:43:19 +0800 "Xu, Shenyuan" wrote: > Inspecting the electron > density map shows that the model does not fit the electron density > well. -- -- Tim Gruene Head of the Centre for X-ray Structure Analysis Faculty of Chemistry University of Vienna Phone: +43-1-4277-70202 GPG Key ID = A46BEE1A To unsubscribe from the CCP4BB list, click the following link: https://www.jiscmail.ac.uk/cgi-bin/WA-JISC.exe?SUBED1=CCP4BB&A=1 This message was issued to members of www.jiscmail.ac.uk/CCP4BB, a mailing list hosted by www.jiscmail.ac.uk, terms & conditions are available at https://www.jiscmail.ac.uk/policyandsecurity/ pgph0PwzhE5yN.pgp Description: OpenPGP digital signature
[ccp4bb] AW: [ccp4bb] Low resolution and high anisotropy
Dear Shenyuan, I have two suggestions: 1. Is your protein a single or a multi-domain protein? If it is a multi-domain protein, you should search with the individual domains. 2. Given the low resolution, high anisotropy and expected large number of molecules in the asymmetric unit, you may have an exceptionally high solvent content, maybe >75%. Is the electron density of all molecules bad, or have a few molecules better electron density? If a few molecules have reasonable density, I would keep those and delete the molecules with bad density. Alternatively, I would start a MR search for only 3 or 4 molecules. Best, Herman Von: CCP4 bulletin board Im Auftrag von Xu, Shenyuan Gesendet: Montag, 24. Oktober 2022 03:43 An: CCP4BB@JISCMAIL.AC.UK Betreff: [ccp4bb] Low resolution and high anisotropy Dear CCP4 community, I have encountered a dataset, which I thought should be easy to solve. The volume of the cell unit seems to be expanded after image 271, which I think is caused by radiation damage. After removing the last few images, the scaled statistics seem good with the resolution set at 3:07 A: d_max d_min #obs #uniq mult. %comp r_mrg r_meas r_pim r_anom cc1/2 cc_ano 91.54 8.33 3086 11972.58 98.44 345.251.30.1010.133 0.0860.145 0.973* -0.148 8.33 6.61 3005 12252.45 99.51 256.527.30.1600.212 0.1380.260 0.923* -0.140 6.61 5.78 2883 11902.42 98.43 92.9 9.40.2780.368 0.2370.461 0.541* -0.165 5.78 5.25 3195 12172.63 99.10 75.6 8.10.2830.369 0.2330.436 0.618* -0.107 5.25 4.87 3192 12062.65 99.42 92.5 8.30.2810.367 0.2320.427 0.856* -0.187 4.87 4.59 3230 12132.66 99.02 118.410.30.2880.378 0.2400.456 0.870* -0.053 4.59 4.36 2797 11562.42 92.93 137.012.20.3090.410 0.2660.496 0.843* -0.239 4.36 4.17 2694 11182.41 93.17 280.216.20.4190.575 0.3920.847 0.491* -0.142 4.17 4.01 3102 11882.61 95.50 133.6 9.90.3920.516 0.3300.622 0.663* -0.268 4.01 3.87 3224 12052.68 98.69 120.4 7.30.4060.539 0.3490.687 0.791* -0.126 3.87 3.75 2939 11812.49 98.91 90.7 6.80.4910.652 0.4250.810 0.691* -0.042 3.75 3.64 1981 10211.94 82.34 87.3 5.00.5570.756 0.5060.885 0.540* -0.018 3.64 3.54 2374 10822.19 89.13 194.611.20.5020.675 0.4470.973 0.625* -0.237 3.54 3.46 2622 11222.34 92.35 310.510.10.4320.585 0.3920.851 0.603* -0.154 3.46 3.38 17399451.84 76.58 48.2 3.60.9301.247 0.8221.542 0.444* -0.046 3.38 3.31 2847 12402.30 98.57 85.4 4.20.6910.930 0.6161.242 0.523* -0.060 3.31 3.24 2838 11532.46 97.88 70.8 3.30.6930.934 0.6201.341 0.362* 0.006 3.24 3.18 3097 12122.56 97.66 71.4 2.50.6920.924 0.6051.173 0.526* 0.044 3.18 3.12 3204 12162.63 99.10 79.6 3.70.6680.886 0.5761.338 0.440* -0.081 3.12 3.07 3059 11722.61 98.16 73.5 2.40.7140.945 0.6121.633 0.362* -0.177 91.50 3.07 57108 232592.46 95.22 138.410.80.3830.513 0.3360.671 0.656* -0.149 I used Mrbump to do the MR, most sequence identities of the starting templates are more than 0.85, and some of them are structures predicted from alpha fold 2. But after refinement (including jelly-body, proSmart, TLC), the best R/Free R stuck at around 0.42/0.45. Inspecting the electron density map shows that the model does not fit the electron density well. The space group is P1, Cell is 58.49 63.97 91.60 91.71 91.86 99.47, should be 5 or 6 molecules in the asymmetric unit. I checked the data quality, it said the data is highly anisotropy. Then I searched the CCP4 forum and used the STARANISO Server and UCLA server, but still cannot improve the refinement. The data statistics after drawing ellipsoidal resolution limits is good: resolution observed redundancycompletenesrmerge i/sigma before/after before/after before/afterbefore/after before/after 7.90 4957 1185 3.5 3.397.3% 92.6%13.0% 12.7%11.9 7.2 5.58 8023 2397 3.0 3.698.6% 98.9%30.3% 11.8% 5.5 7.5 4.5611176 2624 3.3 3.199.1% 97.5%57.1% 18.2% 3.7 6.3 3.9511694 3213 2.9 3.194.0% 99.0%53.1% 27.5% 4.2 4.3 3.5312065 3524 2.6 3.092.8% 99.0%49.8% 45.9% 3.2 2.8