Re: [ccp4bb] should the final model be refined against full datset
Dear Ethan, List, Surely someone must have done this! But I can't recall ever reading an analysis of such a refinement protocol. Does anyone know of relevant reports in the literature? Total statistical cross validation is indeed what we should be doing, but for large structures the computational cost may be significant. In the absence of total statistical cross validation the reported Rfree may be an 'outlier' (with respect to the distribution of the Rfree values that would have been obtained from all disjoined sets). To tackle this, we usually resort to the following ad hoc procedure : At an early stage of the positional refinement, we use a shell script which (a) uses Phil's PDBSET with the NOISE keyword to randomly shift atomic positions, (b) refine the resulting models with each of the different free sets to completion, (c) Calculate the mean of the resulting free R values, (d) Select (once and for all) the free set which is closer to the mean of the Rfree values obtained above. For structures with a small number of reflections, the statistical noise in the 5% sets can be very significant indeed. We have seen differences between Rfree values obtained from different sets reaching up to 4%. Ideally, and instead of PDBSET+REFMAC we should have been using simulated annealing (without positional refinement), but moving continuously between the CNS-XPLOR and CCP4 was too much for my laziness. All the best, Nicholas -- Dr Nicholas M. Glykos, Department of Molecular Biology and Genetics, Democritus University of Thrace, University Campus, Dragana, 68100 Alexandroupolis, Greece, Tel/Fax (office) +302551030620, Ext.77620, Tel (lab) +302551030615, http://utopia.duth.gr/~glykos/
Re: [ccp4bb] should the final model be refined against full datset
For structures with a small number of reflections, the statistical noise in the 5% sets can be very significant indeed. We have seen differences between Rfree values obtained from different sets reaching up to 4%. This is very intriguing indeed! Is there something specific in these structures that Rfree differences depending on the set used reach 4%? NCS? Or the 5% set having less than ~1000-1500 reflections? It would be indeed very interesting if there was a correlation there! A. Ideally, and instead of PDBSET+REFMAC we should have been using simulated annealing (without positional refinement), but moving continuously between the CNS-XPLOR and CCP4 was too much for my laziness. All the best, Nicholas -- Dr Nicholas M. Glykos, Department of Molecular Biology and Genetics, Democritus University of Thrace, University Campus, Dragana, 68100 Alexandroupolis, Greece, Tel/Fax (office) +302551030620, Ext.77620, Tel (lab) +302551030615, http://utopia.duth.gr/~glykos/
Re: [ccp4bb] should the final model be refined against full datset
For structures with a small number of reflections, the statistical noise in the 5% sets can be very significant indeed. We have seen differences between Rfree values obtained from different sets reaching up to 4%. This is very intriguing indeed! Is there something specific in these structures that Rfree differences depending on the set used reach 4%? NCS? Or the 5% set having less than ~1000-1500 reflections? Tassos, by your standards, these structures should have been described as 'tiny' and not small ... ;-) [Yes, significantly less than 1000. In one case the _total_ number of reflections was 5132 reflections (which were, nevertheless, slowly and meticulously measured by a CAD4 one-by-one. These were the days ... :-)) ]. -- Dr Nicholas M. Glykos, Department of Molecular Biology and Genetics, Democritus University of Thrace, University Campus, Dragana, 68100 Alexandroupolis, Greece, Tel/Fax (office) +302551030620, Ext.77620, Tel (lab) +302551030615, http://utopia.duth.gr/~glykos/
Re: [ccp4bb] data processing problem with ice rings
Hi ChenTiantian, the R-factors and I/sigma are bad even at low resolution where the first icering does not influence the results. Thus, the problem with your data processing has little to do with the icerings. I guess that the indexing is not correct. My suggestion: 1) using adxv or a similar display program, note what the inner and outer limits of the ice rings are. These values should be used as parameters for the EXCLUDE_RESOLUTION_RANGE= keywords in XDS.INP, not the provided ones (which are meant for hexagonal ice). 2) start XDS from the INIT step 3) use at least half of your DATA_RANGE as SPOT_RANGE 4) make sure that ORGX and ORGY are correct - mis-indexing is in 90% of the cases due to a wrong origin. In fact, just estimating ORGX and ORGY from the first frame, using adxv or XDS-viewer, seems to do a good job. HTH, Kay Am 20:59, schrieb ChenTiantian: Hi there, I am processing a dataset which has bad ice rings (as you can see in the attach png file). I tried both XDS and imosflm, and got similar results, it seems that adding EXCLUDE_RESOLUTION_RANGE cannot get rid of the effects of the ice rings. the following is part of the CORRECT.LP which is the second attached file, you can find more details there. SUBSET OF INTENSITY DATA WITH SIGNAL/NOISE = -3.0 AS FUNCTION OF RESOLUTION RESOLUTION NUMBER OF REFLECTIONSCOMPLETENESS R-FACTOR R-FACTOR COMPARED I/SIGMA R-meas Rmrgd-F Anomal SigAno Nano LIMIT OBSERVED UNIQUE POSSIBLE OF DATA observed expected Corr 4.24 371525537 5545 99.9% 46.9% 52.7%371502.4850.8%19.4% -28% 0.5135136 3.01 553449002 9840 91.5% 62.7% 65.1%551161.7668.3%48.1% -28% 0.5207760 2.46 84636 12699 12703 100.0% 67.4% 84.7%846341.5573.0%54.2% -19% 0.513 12104 2.13 97910 14743 14987 98.4% 254.5% 199.3%979080.16 276.2% 4899.9% -23% 0.473 14037 1.90 110260 16846 16940 99.4% 299.2% 303.3% 1102450.06 325.0% -99.9% -17% 0.422 15995 1.74 118354 18629 18744 99.4%1062.0% 1043.6% 118317 -0.20 1156.4% -99.9% -13% 0.380 17414 1.61 122958 20193 20331 99.3% 967.5% 1571.1% 1228680.10 1059.7% 987.3%-2% 0.402 18348 1.51 125075 21554 21794 98.9% 838.9% 1355.1% 1249330.08 922.6% 1116.9%-1% 0.402 18977 1.42 72057 17042 23233 73.4% 640.8% 775.3%703910.08 732.5% 826.7%-8% 0.425 10003 total 823746 136245144117 94.5% 166.4% 166.7% 8215620.40 181.1% 296.7% -15% 0.435 119774 Note that I/SIGMA of each resolution shell is 2.5, so how should I do to process the dataset properly? Any suggestion about this super ice rings? Thanks! Tiantian -- Shanghai Institute of Materia Medica, Chinese Academy of Sciences Address: Room 101, 646 Songtao Road, Zhangjiang Hi-Tech Park, Shanghai, 201203 -- Kay Diederichshttp://strucbio.biologie.uni-konstanz.de email: kay.diederi...@uni-konstanz.deTel +49 7531 88 4049 Fax 3183 Fachbereich Biologie, Universität Konstanz, Box 647, D-78457 Konstanz This e-mail is digitally signed. If your e-mail client does not have the necessary capabilities, just ignore the attached signature smime.p7s. smime.p7s Description: S/MIME Kryptografische Unterschrift
Re: [ccp4bb] Insufficient virtual memory
Hi Ian, compiling on your 32bit machine gave you a 32bit binary, so your 12GB RAM cannot be used! HTH, Kay Am 20:59, schrieb Ian Tickle: Hello all, some Fortran developer out there must know the answer to this one. I'm getting a forrtl: severe (41): insufficient virtual memory error when allocating dynamic memory from a F95 program compiled with Intel Fortran v11.1.059. The program was compiled on an old ia-32 Linux box with 1Gb RAM + 2Gb swap (I only have one Intel license to compile on this machine), but I'm running it on a brand new x86-64 box with 12Gb RAM + 8Gb swap. This should be ample: the program's maximum total memory requirement (code + static data + dynamic data) should be no more than 3Gb. My question is: what do I have to do to make it work? According to the ifort man page I need to specify -mcmodel=medium -shared-intel. It says: If your program has COMMON blocks and local data with a total size smaller than 2GB -mcmodel=small is sufficient. COMMONs larger than 2GB require mcmodel=medium or -mcmodel=large. Allocation of memory larger than 2GB can be done with any setting of -mcmodel. I'm a bit confused about the difference here between COMMONS 2Gb (which I don't have) and allocation of memory 2Gb (which I assume I do). When I try setting -mcmodel=medium (and -shared-intel) I get ifort: command line warning #10148: option '-mcmodel' not supported. Is this telling me that I have to compile on the 64-bit machine? Whatever happened to cross-compilation? All suggestions greatly appreciated! -- Ian -- Kay Diederichshttp://strucbio.biologie.uni-konstanz.de email: kay.diederi...@uni-konstanz.deTel +49 7531 88 4049 Fax 3183 Fachbereich Biologie, Universität Konstanz, Box 647, D-78457 Konstanz This e-mail is digitally signed. If your e-mail client does not have the necessary capabilities, just ignore the attached signature smime.p7s. smime.p7s Description: S/MIME Kryptografische Unterschrift
Re: [ccp4bb] data processing problem with ice rings
Hi I'd agree with Kay here - I would think that the original indexing is incorrect. One thing I notice on the original image as posted - there's a red cross on it - if that's supposed to mark the beam position, I think it's about 4mm or so away from the true position. So - (1) check the beam position carefully (it may be wrong in the image header) (2) after indexing, make sure that the predictions match the spot positions (3) if the predictions don't match the spot positions, don't try to integrate - find out what's wrong (wrong wavelength, beam position, distance???). If you can't work it out, ask one of the experts to look at a sample of your original images (iMosflm ask Andrew or me, XDS ask Kay, HKL Wladek or ZO...). (4) If the predictions do match the spot positions, integrate the dataset in P1 (i.e. triclinic) and see what Pointless suggests as the symmetry. You may just be trying to impose too much symmetry. If you can't work out what the issue is, ask an expert to help directly - we're all happy to help out! (5) Worry about the ice rings after you've sorted out the above problems, not before. HTH On 15 Oct 2011, at 12:09, Kay Diederichs wrote: Hi ChenTiantian, the R-factors and I/sigma are bad even at low resolution where the first icering does not influence the results. Thus, the problem with your data processing has little to do with the icerings. I guess that the indexing is not correct. My suggestion: 1) using adxv or a similar display program, note what the inner and outer limits of the ice rings are. These values should be used as parameters for the EXCLUDE_RESOLUTION_RANGE= keywords in XDS.INP, not the provided ones (which are meant for hexagonal ice). 2) start XDS from the INIT step 3) use at least half of your DATA_RANGE as SPOT_RANGE 4) make sure that ORGX and ORGY are correct - mis-indexing is in 90% of the cases due to a wrong origin. In fact, just estimating ORGX and ORGY from the first frame, using adxv or XDS-viewer, seems to do a good job. HTH, Kay Am 20:59, schrieb ChenTiantian: Hi there, I am processing a dataset which has bad ice rings (as you can see in the attach png file). I tried both XDS and imosflm, and got similar results, it seems that adding EXCLUDE_RESOLUTION_RANGE cannot get rid of the effects of the ice rings. the following is part of the CORRECT.LP which is the second attached file, you can find more details there. SUBSET OF INTENSITY DATA WITH SIGNAL/NOISE = -3.0 AS FUNCTION OF RESOLUTION RESOLUTION NUMBER OF REFLECTIONSCOMPLETENESS R-FACTOR R-FACTOR COMPARED I/SIGMA R-meas Rmrgd-F Anomal SigAno Nano LIMIT OBSERVED UNIQUE POSSIBLE OF DATA observed expected Corr 4.24 371525537 5545 99.9% 46.9% 52.7%371502.4850.8%19.4% -28% 0.5135136 3.01 553449002 9840 91.5% 62.7% 65.1%551161.7668.3%48.1% -28% 0.5207760 2.46 84636 12699 12703 100.0% 67.4% 84.7%846341.5573.0%54.2% -19% 0.513 12104 2.13 97910 14743 14987 98.4% 254.5% 199.3%979080.16 276.2% 4899.9% -23% 0.473 14037 1.90 110260 16846 16940 99.4% 299.2% 303.3% 1102450.06 325.0% -99.9% -17% 0.422 15995 1.74 118354 18629 18744 99.4%1062.0% 1043.6% 118317 -0.20 1156.4% -99.9% -13% 0.380 17414 1.61 122958 20193 20331 99.3% 967.5% 1571.1% 1228680.10 1059.7% 987.3%-2% 0.402 18348 1.51 125075 21554 21794 98.9% 838.9% 1355.1% 1249330.08 922.6% 1116.9%-1% 0.402 18977 1.42 72057 17042 23233 73.4% 640.8% 775.3%703910.08 732.5% 826.7%-8% 0.425 10003 total 823746 136245144117 94.5% 166.4% 166.7% 8215620.40 181.1% 296.7% -15% 0.435 119774 Note that I/SIGMA of each resolution shell is 2.5, so how should I do to process the dataset properly? Any suggestion about this super ice rings? Thanks! Tiantian -- Shanghai Institute of Materia Medica, Chinese Academy of Sciences Address: Room 101, 646 Songtao Road, Zhangjiang Hi-Tech Park, Shanghai, 201203 -- Kay Diederichshttp://strucbio.biologie.uni-konstanz.de email: kay.diederi...@uni-konstanz.deTel +49 7531 88 4049 Fax 3183 Fachbereich Biologie, Universität Konstanz, Box 647, D-78457 Konstanz This e-mail is digitally signed. If your e-mail client does not have the necessary capabilities, just ignore the attached signature smime.p7s. Harry -- Dr Harry Powell, MRC Laboratory of Molecular Biology, MRC Centre, Hills Road, Cambridge, CB2 0QH
Re: [ccp4bb] How to calculate energy?
This should all be in the CNS output file for each cycle. Boaz Boaz Shaanan, Ph.D. Dept. of Life Sciences Ben-Gurion University of the Negev Beer-Sheva 84105 Israel E-mail: bshaa...@bgu.ac.il Phone: 972-8-647-2220Skype: boaz.shaanan Fax: 972-8-647-2992 or 972-8-646-1710 From: CCP4 bulletin board [CCP4BB@JISCMAIL.AC.UK] on behalf of Huayue Li [lihua...@naver.com] Sent: Saturday, October 15, 2011 5:41 AM To: CCP4BB@JISCMAIL.AC.UK Subject: [ccp4bb] How to calculate energy? Dear all, I obtained 20 peptide models (with lowest energy) calculated by CNS program. Now I want to make a table for structure statistics, but I don't know how to calculate Ebond Eangle Eimproper Evdw ENOE Ecdih Etotal , r.m.s. deviation from experimental constraints, and r.m.s. deviations from idealized geometry. Where can I get these information? Just from the peptide pdb file output by CNS, or using another software? And what is idealized geometry? Thanks! Huayue Li, Ph. D College of Pharmacy Pusan National University Geumjeong-gu, Jangjeon-dong Busan 609-735, Korea Tel: 82-51-510-2185
Re: [ccp4bb] Akta Prime
Dear Mike, At Evotec (a UK site), we gave up using GE to service our GE instruments (!) due to problems with their bureaucracy. Agilent offer service contracts for Aktas that are competitively priced and are a work-alike in our experience. For call-outs, they sub-contract the usual GE engineers that you would normally see to perform maintenance, and do this with their usual high level of professionalism. I think it’s a bit of a shame (for GE) that we have to use a third party like this to organise the preventative maintenance visits etc., but it works very well for us. Yours, Mark On 12 October 2011 19:28, Michael Colaneri colane...@gmail.com wrote: Dear all, We have an AktaPrime and GE Lifesciences stop servicing these instruments because they are getting old. Does anyone know of a third party company that gives contracts to maintain these instruments? Thank you. Mike Colaneri
[ccp4bb] small 3ml Superdex column performance
Hi everyone: I was hoping I could get your opinion on the performance of the small approx. 3ml gel filtration columns from GE Healthcare. We currently have an Akta fplc and we would like to do small runs using small volumes of sample. As far as I know we have two options: 1. use the Superdex 15/150 columns these appear to have a slightly lower resolution (the no. of theoretical plates is about 25000 m-1) these apparently can be directly hooked up to our Akta fplc 2. use Superdex PC 3.2 these have somewhat higher resolution (the no of theoretical plates is 3 m-1) these can only be hooked up using a special Precision Holder equipped with titanium end fittings (which I have heard clog easily?) What is your experience with these columns? Which option of the two do you recommend (cost is not a huge issue, but resolution and overall performance is)? Have you seen significant band broadening when using these small columns with a regular fplc (rather than Akta's microFPLC)? I would greatly appreciate your comments! Many thanks... Bests, Alex .
Re: [ccp4bb] data processing problem with ice rings
Hi, I agree with other people. You must have a wrong index here. Can you tell us what is the unit cell for this crystal from your determination? I can see very close spots in the high resolution shell from your image, which are overlapped into one spot in the low resolution shell. Try to use other frames to do the indexing. If it is still not working, it might be easier to collect another dataset with better crystal alignment. Best, Nian On Fri, Oct 14, 2011 at 12:12 AM, ChenTiantian chentiantian2...@gmail.comwrote: Hi there, I am processing a dataset which has bad ice rings (as you can see in the attach png file). I tried both XDS and imosflm, and got similar results, it seems that adding EXCLUDE_RESOLUTION_RANGE cannot get rid of the effects of the ice rings. the following is part of the CORRECT.LP which is the second attached file, you can find more details there. SUBSET OF INTENSITY DATA WITH SIGNAL/NOISE = -3.0 AS FUNCTION OF RESOLUTION RESOLUTION NUMBER OF REFLECTIONSCOMPLETENESS R-FACTOR R-FACTOR COMPARED I/SIGMA R-meas Rmrgd-F Anomal SigAno Nano LIMIT OBSERVED UNIQUE POSSIBLE OF DATA observed expected Corr 4.24 371525537 5545 99.9% 46.9% 52.7% 371502.4850.8%19.4% -28% 0.5135136 3.01 553449002 9840 91.5% 62.7% 65.1% 551161.7668.3%48.1% -28% 0.5207760 2.46 84636 12699 12703 100.0% 67.4% 84.7% 846341.5573.0%54.2% -19% 0.513 12104 2.13 97910 14743 14987 98.4% 254.5%199.3% 979080.16 276.2% 4899.9% -23% 0.473 14037 1.90 110260 16846 16940 99.4% 299.2%303.3% 1102450.06 325.0% -99.9% -17% 0.422 15995 1.74 118354 18629 18744 99.4%1062.0% 1043.6% 118317 -0.20 1156.4% -99.9% -13% 0.380 17414 1.61 122958 20193 20331 99.3% 967.5% 1571.1% 1228680.10 1059.7% 987.3%-2% 0.402 18348 1.51 125075 21554 21794 98.9% 838.9% 1355.1% 1249330.08 922.6% 1116.9%-1% 0.402 18977 1.42 72057 17042 23233 73.4% 640.8% 775.3%703910.08 732.5% 826.7%-8% 0.425 10003 total 823746 136245144117 94.5% 166.4%166.7% 8215620.40 181.1% 296.7% -15% 0.435 119774 Note that I/SIGMA of each resolution shell is 2.5, so how should I do to process the dataset properly? Any suggestion about this super ice rings? Thanks! Tiantian -- Shanghai Institute of Materia Medica, Chinese Academy of Sciences Address: Room 101, 646 Songtao Road, Zhangjiang Hi-Tech Park, Shanghai, 201203
Re: [ccp4bb] small 3ml Superdex column performance
Hi, You're probably referring to the Superdex 5/150 Tricorn column, with working volume of 3ml, and not the 15/150? Those columns work quite nicely for small sample volumes. For analytical runs 15-25ul injection is pretty nice. The PC columns are originally designed to be used with the SMART system, which by the way used to be one of the best analytical products for macromolecules -- optimized path length, one-volume pumps suitable for directly running typical protocols, etc. etc. and for its time the OS was also damn good (OS/2). Sadly, GE did not come up with any direct replacement for this machine, as far as I can tell. AKTA is not optimal for analytical runs - tubing is too long, etc. If you have an old HPLC system moldering in a corner I recommend either of these columns mounted directly in front of the detector. In our current setup the entire portion of the HPLC that is responsible for column selection and heating and so on is bypassed, so there are literally ~4-5 mm of (the thinnest available PPEK) tubing in between the injection valve and the column, and in between the column and the detector. Autosampler is a very helpful feature, esp. when analyzing fractions output from previous step, and as long as the column isn't clogged the run is 8-12 minutes (depending on buffer composition). Now, Agilent software for HPLC is absolutely horrible for this kind of work but it suffices. Artem P.S. for lower protein quantities don't forget to record the A210, in addition to A280 and A260. On Sat, Oct 15, 2011 at 5:04 PM, Alexandra Deaconescu deac...@brandeis.eduwrote: Hi everyone: I was hoping I could get your opinion on the performance of the small approx. 3ml gel filtration columns from GE Healthcare. We currently have an Akta fplc and we would like to do small runs using small volumes of sample. As far as I know we have two options: 1. use the Superdex 15/150 columns these appear to have a slightly lower resolution (the no. of theoretical plates is about 25000 m-1) these apparently can be directly hooked up to our Akta fplc 2. use Superdex PC 3.2 these have somewhat higher resolution (the no of theoretical plates is 3 m-1) these can only be hooked up using a special Precision Holder equipped with titanium end fittings (which I have heard clog easily?) What is your experience with these columns? Which option of the two do you recommend (cost is not a huge issue, but resolution and overall performance is)? Have you seen significant band broadening when using these small columns with a regular fplc (rather than Akta's microFPLC)? I would greatly appreciate your comments! Many thanks... Bests, Alex .
