From james holton Sent Wednesday, September 17, 2003 4:58 pm To beamline-support@lbl.gov Cc Chris Nielsen Bcc Subject What is the best wavelength to use? Something that might help answer a commonly-asked user question: I recently tried to empirically look into the "best" wavelength to collect at. Although 1.1A represents the best signal into a photodiode on 8.x.x, but it appears that the ADSC detectors have a different energy response curve. I processed the same wedge from the same crystal at a wide range of energies (with exposure times roughly normalized to the photon flux and distance adjusted to have the same resolution on the edge of the detector). energy dist photons dose(mJ) absorbed I/sd fom 6000 300 9.2e+12 8.85 1.66 25.6 0.060 7000 350 7.4e+12 8.29 1.00 27.8 0.108 8000 400 6.0e+12 7.67 0.62 29.9 0.186 9000 450 5.2e+12 7.57 0.43 27.6 0.247 10000 500 4.9e+12 7.84 0.33 34.1 0.405 11000 550 4.6e+12 8.08 0.25 44.1 0.675 13000 650 4.2e+12 8.84 0.17 40.2 0.917 14000 700 4.3e+12 9.70 0.15 39.0 1.000 15000 750 4.6e+12 10.95 0.14 30.1 0.826 Where: energy - is the photon energy dist - is the xtal-to-detector distance photons - is the total number of photons used in 10 exposures dose - is the amount of energy in those photons (in milliJoules) absorbed - is the amount of energy expected to have been absorbed in 100um of protein crystal I/sd - is the final, overall signal-to-noise ratio of the data set fom - normalized (I/sd)/(absorbed dose) The "figure of merit" in the last column is probably the parameter you want to maximize in a damage-limited run. The absorbed dose is calculated from the expected absorption of 100um of protein crystal. This would imply that the "best wavelength" is actually around 13-14keV when using an ADSC detector. So, at higher energies, you get less flux, but you do less damage for a given signal/noise ratio. Questions/comments (especially on my math) are welcome. -James From james holton Sent Monday, September 29, 2003 1:15 am To beamline-support@lbl.gov Cc tom@ucxray.berkeley.edu Bcc Subject the best exposure time Another one of those questions I get asked a lot and don't have a good answer to is "What should I use for an exposure time?" During 2-bunch I did an experiment collecting SAD data from one of Corrie's Gd-lysozyme crystals at a variety of different exposures. The flux was low here, and the total dose far all 18 runs combined was 5x10^9 photons/um^2, which I think is still comfortably below the most pessimistic "damage limits". The results are: exposure frames Rmerge Ranom I/sd Mult MADPPH MPDFPH FOM FOMDM CC 0.1s 100 0.100 0.060 13.3 7.5 5.29 28.89 0.138 0.507 0.245 0.1s 200 0.107 0.054 19.0 15.1 6.35 32.67 0.229 0.615 0.364 0.1s 300 0.109 0.052 23.5 22.6 6.58 34.17 0.265 0.676 0.374 0.1s 400 0.110 0.050 27.2 30.2 6.71 35.39 0.289 0.641 0.447 0.1s 500 0.110 0.049 30.5 37.8 6.83 35.74 0.303 0.680 0.456 0.1s 600 0.111 0.049 33.5 45.3 6.65 36.19 0.313 0.647 0.463 0.1s 700 0.111 0.048 36.2 53.0 6.92 36.68 0.325 0.656 0.459 0.1s 800 0.111 0.048 38.7 60.5 6.93 36.92 0.333 0.701 0.455 0.1s 900 0.112 0.048 41.1 68.1 7.24 37.02 0.339 0.705 0.449 0.1s 1000 0.112 0.047 43.3 75.7 7.23 37.11 0.343 0.711 0.452 1.0s 100 0.047 0.047 33.3 7.6 7.61 37.93 0.366 0.726 0.441 1.0s 200 0.050 0.045 48.1 15.1 8.02 39.44 0.409 0.708 0.456 2.0s 100 0.041 0.046 40.3 7.5 8.08 38.92 0.399 0.761 0.454 1.0s 100 0.047 0.047 33.3 7.6 7.61 37.93 0.366 0.726 0.441 1.0s 200 0.050 0.045 48.1 15.1 8.02 39.44 0.409 0.708 0.456 1.0s 300 0.051 0.045 59.0 22.7 8.58 39.81 0.424 0.687 0.464 1.0s 400 0.052 0.044 67.9 30.3 8.40 39.95 0.433 0.745 0.459 2.0s 200 0.043 0.045 57.7 15.1 8.51 39.93 0.432 0.688 0.475 4.0s 100 0.036 0.044 46.0 7.5 8.08 39.70 0.413 0.655 0.395 Where: exposure - is the (dose mode) exposure time frames - number of frames involved in the data set Rmerge - sum(|I-|)/sum(I) Ranom - sum( - )/sum( + ) I/sd - signal-to-noise in intensity measurement Mult - mean observation redundancy MADPPH - maximum anomalous difference Patterson peak height MPDFPH - maximum phased anomalous difference Fourier peak height FOM - figure of merit (mlphare) FOMDM - figure of merit after solvent flattening CC - correlation coefficient of the solvent-flattened map to the known model All the 0.1s runs are subsets of the data used for the 1000 image run. Now, the first thing I thought was interesting about this is that 1000 0.1s exposures are just as good as 100 1s exposures. The spots on the 0.1s frames were REALLY faint. I'm amazed that mosflm could even hold onto them. I think this speaks to the low read noise in our detectors. This also suggests that calculating a collection strategy could be a complete waste of time. You might as well divide the same total dose over 180 degrees. The time taken calculating the strategy hardly adds up to the summed detector overhead of a 180-degree run. Also, if there really is no signal penalty to dividing the data over many frames, the optimum compromise dose for the singal/damage ratio can be found long after data are collected. What I also found striking about this is that the overall quality of the experimental map (CC) reaches a plateau at very low dose. I think this suggests an upper limit to the quality of SAD phases. Doubling the signal-to-noise ratio from 30 to 60 does not significantly improve the FOM or the agreement between map and model. Now, at 11111eV, Gd is supposed to have 8.3 anomalous electrons. If I look at the refined mlphare output, I see three Gd sites with an average contribution of 3 anomalous electrons each. Since I'm pretty sure I got the structure factor scaling right in "truncate", I suppose the 3 electrons could be due to <100% occupied Gd sites. Using the Crick-Magdoff equation: = (f"/fP)*sqrt(2*NH/NP) delta-I - the average intensity difference due to anomalous scattering f" - number of anomalous electrons in each heavy atom fP - number of electrons in each protein atom (7) NH - number of heavy atoms (3) NP - number of protein atoms (~1100) we get an expectation of ~1% difference per anomalous electron. So, our expected signal here is about 3%. Now, when I/sd is 30, the noise in difference data should be a factor of sqrt(2) higher than the noise in the data. So, at I/sd=30, we can expect an anomalous difference noise level of sqrt(2)/30 = 4.7%. Which means a signal-to-noise ratio of about 0.7 for difference data is all that is required to get the best map. Kind of hard to believe? What is potentially useful about all this is that you might be able to calculate the needed exposure time for a given heavy-atom situation: 1) given the protein size and expected number of heavy atoms, you can calculate the needed signal/noise to get a difference signal/noise of 0.7: I/sd = 1.3 *sqrt(MW/NH)/f" where MW is the molecular weight in Da. 2) for a given exposure, you can measure the peak height/noise level of an "average"-looking spot in mosflm, and this should be comparable to the I/sd in the final data for minimal-strategy data collection. I'll have to do some more experiments to see if this theory really works or not. As usual, questions and comments (especially about the math) are welcome... -James From james holton Sent Tuesday, November 11, 2003 10:24 am To "Deacon, Ashley" Cc beamline-support@lbl.gov Bcc Subject Re: Low energy limit Cool, congratulations! So, how did you do your collection strategy? I've heard many rumors about needing special scaling programs and multiaxis goniostats, is any of it true? How easy was it to trace the chain? So many questions... We're getting more and more people asking about the "right" way to do Sulfur anomalous and I've never had anything too helpful to tell them other than "are you SURE you can't get Selenium in there?" -James "Deacon, Ashley" wrote: > FYI > > We solved the first Sulfur-SAD dataset we colelcted on friday night. > We're just analyzing all the data from the others, but at least one > of them looks promising. > > We got 50Gb of data or so during the time I think!! > > Ashley. > > -----Original Message----- > From: JMHolton@lbl.gov [mailto:JMHolton@lbl.gov] > Sent: Friday, October 31, 2003 9:28 PM > To: Deacon, Ashley > Subject: Re: Low energy limit > > > > ----- Original Message ----- > From: "Deacon, Ashley" > Date: Thursday, October 30, 2003 10:56 pm > Subject: Low energy limit > > > Hi James, > > > > What is the low energy limit on 8.3.1?? > > The main thing stopping low-energy light at 8.3.1 > is air. Our peak flux is actually around 3-4 keV, > but air knocks the maximum up to 11 keV. Still, > we have useful flux down to 6 keV, 5 is possible, > but then you have to start worrying about Si(333) > contamination. Let me know if you want to go down > that far, and I can show you how to do it. > > > > > We are interested in pursuing sulfur phasing > > next week on 8.3.1 and were wondering how low > > we can go... we were thnking to work close to > the > > iron edge... > > > > We have a protein with 8 S per 150 amino acids > > molecule and we cannot get Se protein. > Fortunately > > it diffracts to 1.25A or so!! It is also high > symmetry > > P622... > > > > How is flux at iron edge if you can get there?? > > We get 1.1x10^11 photons/s at 7keV. The > absorption of protein here is 5x higher than at > 12keV. A rough calculation suggests you can > expect your crystals to last 3-4 hours. > > Hope this is helpful! > > -James > > > > > From James Holton Sent Saturday, December 13, 2003 2:33 pm To beamline-support@lbl.gov Cc Tom_Scarvie@lbl.gov Bcc Subject Re: elongation of spots The elongation of spots on the detector is the product of the beam divergence and the detector distance. If the detector distance is less than ~200mm, you won't see it because 3mrad*200mm = 0.6mm = 6 pixels, which is comparable to the natural size of most spots. But, this becomes VERY noticable at long detector distances. (3mrad*1m = 3mm) MOST USERS ARE NOT AWARE OF THIS! Home-lab x-ray generators do not generally have adjustable divergence. (in fact, they usually focus the beam on the detector surface). So, I think we can expect users to not understand this "feature" of our beamlines. I have even heard users describe throwing out perfectly good samples because the spots "started looking like little frowny faces". (This is bad.) Now, I'm certainly not an expert in optics. Alastair MacDowell, Howard Padmore or Malcom Howells would be the best people to ask about the expected effect of electron beam changes on detector spot size. However, my understanding is that any enlargement of the source due to "weirdness" in the storage ring would manifest itself as a proportional enlargement of the image of the source at the beamline's focus (the pinhole). Any changes in the shape of the source would just be apertured off, and produce a drop in brightness with no real change in spot shape at the detector. In the near future, I will change 8.3.1's control rules so that the spot size will automatically adjust to a "sane" value at long detector distances. But, until we are all doing this, user training is the only answer. (This is really bad.) -James -----Original Message----- >From: "Jonathan Spear" >Sent: 12/11/03 12:07:34 AM >To: "beamline-support@LBL.gov" >Cc: "Tom_Scarvie@lbl.gov", "DBRichardson@lbl.gov" >Subject: Horizontal elongation of diffraction spots > >Martin, the user on 8.2.2 called at 11PM complaining about horizontal >elongation of all of the diffraction spots on his images. He asked me if I >thought it would help for him to swap out the collimator. I suggested >instead that he first try narrowing the slits (horizontal convergence in >DCS) by a factor of two to see if this would help. A few minutes later, >Martin's colleague, Greg, called and reported that they reduced the >horizontal convergence from 2.0 murad to 0.8 murad, and increased the >exposure time from 12 seconds to 20 seconds. The result was that their >spots were resolvable, though still elongated. > >Because Ina, the user on 5.0.3, also had mentioned this apparent problem >this afternoon, I thought that it might be an issue of positional >instability of the storage ring, or some such phenomenon. I called the >control room at X4969 and spoke with Dave Richardson, suggesting that there >might be some spatial instability in the electrons. Dave reported that >there have been some recent adjustments in the orbital control system, and >that he would report our observations to Tom Scarvie of Physics tomorrow. > >I made a followup call to our other beamlines, to see if anyone else was >having the same problems. The users on 8.3.1 and 5.0.2 said that they had >no problem differentiating between spots, although their unit cells were not >very large. This information suggests that there is NOT really a problem >with the ALS. > >Perhaps someone can look at their images, from about 10PM Wednesday until >midnight, to see if the spots look normal? > > >-Jon Spear > From james holton Sent Thursday, January 29, 2004 12:01 pm To beamline-support@lbl.gov Cc Bcc Subject [Fwd: data update] Good news from a user attached. I thought it was particularly interesting that this was the fifth time now I have heard of someone solving a structure because they reduced their data set to "early" data only. Radiation damage sucks... Spread the word, O ye support personel! -James ----- Original Message ----- From Joy Huffman Date Thu, 29 Jan 2004 10:43:22 -0800 (PST) To JMHolton@lbl.gov Subject data update Hi James, Maybe you remember helping me with my MAD data collection last Friday night. I was trying to get phases for a brominated DNA structure, and we collected several different datasets, and you were running Elves and made some maps, etc. It didn't seem to work at the time. We collected first a single dataset at the peak (/data/mcfuser/scripps/joy/jan04/br3) with 1 sec exposures. This one didn't give phases. Then we collected a full MAD dataset with 5 sec exposures, 1 degree wedge, inverse beam, the whole kit and kaboodle. My space group is F432, so we collected ~55 degrees. When I ran Solve with all of the data, the FOM was ~.33, which was better than usual, but the maps still weren't interpretable (by me anyways). I ran Solve with only 30 degrees, so the data are less complete and redundant, but the crystals were exposed to less radiation, and the FOM was 0.43, and the maps are clearly interpretable with FOM=0.57 after solvent flattening. Yeah! Anyway, I'm just letting you know that the setup for collecting 1 degree wedges at your beamline is wonderful, and your help with data collection strategies is very much appreciated. I would otherwise have assumed more data was always better and probably spent a lot more time trying to phase my data. Thanks again, Joy *********************************** Joy L. Huffman, Ph.D. The Scripps Research Institute Mailcode MB4, Room MB104 10550 N. Torrey Pines Rd. La Jolla, CA 92037 phone: 858-784-8597 fax: 858-784-2289 *********************************** From GMeigs@lbl.gov Sent Thursday, February 5, 2004 7:46 am To GMeigs@lbl.gov Cc Bcc Subject focusing beam on detector ----- Original Message ----- From james holton Date Wed, 04 Feb 2004 18:37:26 -0800 To beamline-support@lbl.gov Subject Re: saturday start up Since the answer to this question is pretty much the same on all the beamlines... 1) logistically: This kind of move really does need to be arranged more than -30min in advance. It takes at least 2 hours to set up and 2 hours to tear down from a refocus (on a good day). SIBYLS will be able to do this much more easily (right JJ?). At the moment, I would say 8.3.1 cannot do a refocusing move for someone who does not have a weekday daytime changeover to the next user. 2) Physically: Refocusing with a crystal that is smaller than the focused beam (~100um) is useless. There is no optical difference between having the focus on the detector at 1000mm with a 100um crystal and having the focus at the sample with the divergence set to 0.1mrad. Basically, the crystal acts as a "slit" which is 0.1mm wide and 1000mm from the focus (0.1mrad), so the flux and damage rate are the same. The only time refocusing is useful is when you have a crystal that is bigger than the beam focus size. This is why we always use a bigger pinhole when we refocus. In this case, putting the focus on the detector allows the beam to spread out over the large sample area. The crystal lives longer for the same flux. You get the same flux you would have with the divergence set to the pinhole width: (I.E. 200um at 1m is 0.2mrad) - So if someone asks to refocus the beam, ask them how big their crystal is. (If they say it's "big" ask them to convert that to metric units) - If the crystals really are ~200um or bigger, then they have a case for refocusing (1/4 exposure time). - If refocusing is impractical because of scheduling, then the next best thing is chopping down the slits and shooting different parts of the big crystal. A good rule of thumb for the slit width is: Hdiv = 0.3*1000/distance Where the distance is in mm and the divergence will be in mrad. ABDauz@lbl.gov wrote: > > 831 user called and asking a way to focus back the beam because she > have a very low resolution diffracting crystal and having detector set > to 800 mm distance to catch the low resolution spots. I wasnt sure > about this so I called George M and asked him to call 831 user for > help, thanks George. > > -Azer From james holton Sent Sunday, February 22, 2004 12:53 pm To beamline-support@lbl.gov Cc Bcc Subject minimum exposure time I did an experiment to address the question: "What is the minimum exposure time we can use here?" The answer appears to be "0.1s" For some reason I cannot yet identify, I see a precipitous drop-off in the signal-to-noise of a lysozyme data set below 0.1s exposure time. I took 10 data sets of 100 images each using the same total delivered dose (2.1x10^11 photons) with different exposure times ranging from 0.02s to 4s. That is, I used foils and slits to attenuate the beam so that the flux into the photodiode was inversely proportional to the exposure time (within a precision of <5%) for each data set. That way, I get the same number of photons per data set, and the exposure time is the only variable. The total, overall delivered dose to the crystal was < 3x10^7 ph/um^2, which is ~1/4,000 the "Henderson dose" of 1x10^11 ph/s expected to significantly damage a protein crystal, so I don't think the crystal was damaged in this run. Nevertheless, the short exposures were done first. I processed all these data sets and scaled them against each other. The average scale factor (the number you need to multiply the data by to make it as close as possible to the other data) was fairly constant (within ~4% of 1.0), which suggests that the dose delivered in each data set was fairly constant. However, the signal-to-noise ratio (I/sd) dies off rapidly below 0.1s. Graph attached. A drop in I/sd for a constant dose is definitely something to worry about because it reflects loss of usable data for a given amount of damage to the sample. I think the drop we see here must be coming from jitter in the fast shutter, jitter in the spindle(?) or high-frequency (>20 Hz) components in the beam intensity. NM laser advertizes "Switching speed under 2 msec", and "Exposures under 4 msec" for the LS055 (Which I think is the model we have). They never said anything about jitter. We do have high-frequency components in the beam, but I'm not sure how high, exactly. I admit we havn't looked at this in a long time. In the early days, there wre some ~7 Hz components from shaking of water hoses in the mono, but those are gone now. Might be interesting to try this again when we replace the water hoses in May. I guess the comforting thing is that the I/sd is pretty steady right up until 0.1s, so we can be confident nothing screwy is going on for those exposure times. Anyway, hope this answers at least one common user question. Questions/comments are always welcome. -James File: exposure_vs_dose.gif From james holton Sent Monday, April 26, 2004 7:51 pm To frasm@ucalgary.ca Cc GMeigs@lbl.gov , JTanamachi@lbl.gov Bcc Subject Re: data collection on the 400 A cell Marie Fraser wrote: > Hello, James, > Thank you for collecting the data on the SCS mutant with the > 400 A cell. Karla told me that you had to help them on that. > Normally, I keep the detector in as close as possible, then fight > with choosing a spot size and background. So it was interesting > to see that you chose to move the detector back. (If you see the > end-of-run forms, you may gather that I am the person who > likes the 3x3 detector.) Do you like it enough to give me US$1M so I can buy you one? ;) > Your way made better use of the low > divergence of the beam, and I didn't have to cycle through the > processing as much to pick a spot size. I try to err on the side of resolved spots. > One thing I neglected to mention in the end-of-run form > was that we sometimes had trouble moving the detector in > for the first image after putting up a new crystal. I wasn't > getting the command to shut the door, but Stephanie did > wonder if it was due to the contact for the table that the > liquid nitrogen sits on. Karla and Joanne didn't have the > problem, only I did. I am showing 4 following errors for detector_z at: Apr 21 15:38:29 2004 600.0 Apr 21 15:38:29 2004 600.0 Apr 22 16:18:01 2004 600.0 Apr 24 12:22:43 2004 600.0 Since all of these were trying to go to 600mm, I assume they were "park" moves. Wouldn't expect you to have noticed that. We also had 12 following errors on detector twotheta, which will kill detector_z moves (to prevent disasters). This has been a real pain in the ass lately. I've got our motor control expert looking into this now. > Are you able to use Elves to process data with the 2-theta > swing out? Yup, the processing is here in: /home/mcfuser/alberta/apr20/MEFv1/MEFv1a and at: http://bl831.als.lbl.gov/~jamesh/pickup/MEFv1.tar.gz > I tried, but gave up in frustration on Friday and > processed with denzo/scalepack on Saturday. I think that what > I needed to do was take the orientation matrix from the low > resolution data collection. Instead, I tried to pick spots with > mosflm and I didn't have the detector settings right so mosflm > saw nothing. You need to use a twotheta angle that is the negative of what is in the image headers. Wedger Elves know this, and should do it right. At least, it worked for me. But then again, twotheta data are always a little tricky, mainly because it's usually so long between when I do them, that I forget everything I learned last time. Still, it's tractable. hope it turns out to be useful. In the future, I'll see what I can do about using our SIBYLS connections to swap time with them for large-unit-cell projets. > I had read your Elves paper, and your commentary > in an online journal - unfortunately, I need more mosflm > experience to do the tricky things. > Many thanks, Marie Good luck! -James From james holton Sent Wednesday, April 28, 2004 12:53 pm To erica Cc BTGreensmith@lbl.gov , GMeigs@lbl.gov , JJPlecs@lbl.gov Bcc Subject Re: questions about PT on 8.3.1 Yes, we do have Pt in our mirrors, but it's covered with Rhodium to minimize the dip in flux across the Pt edges (a percent or so). The impact of the Pt coating on Pt fluorescence scans is that you would get an artifical _reduction_ in the fluorescence signal, because the incident flux suddenly drops. This would weaken the apparent size of the edge. From your scans, it looks to mee like you did see some of this effect, but the white line still shone through. I'm sorry to say that the destruction of your anomalous signal in the first 20 frames is not really suprizing to me. With 5s exposures, you would expect to loose most of the phasing signal in 60 images. It's possible for certain "tough" crystals to last a lot longer, but I usually see "site death" in less than 300s. Depending on what you use to measure the anomalous signal, it will drop under the noise sooner or later than that. What did you use to measure your anomalous signal? The temperature of the cryostream is something that people do wonder about with respect to damage, but few have measured it. To my knowledge, there have been two studies exploring damage rates at temperatures below 90K, with conflicting results. (Actually, if Wladek has any references for controlled studies of the impact of sample temperature on crystal lifetime I'd be really interested in hearing about them.) Most people seem to point to the glass transition work done by Petsko in the late 70s that anything below ~120K should "freeze out" the free radicals that cause nearly all the damage at room temperature. Still, I do worry about this. It's possible for the glass transition temperature to depend on the cryosolvent used. Not much is known about this. One of the reasons I have a coaxial cryostream is that it seems to make the sample ~20 degrees colder than if you have the stream at an angle. This is based on measurements I made with a tiny thermocouple mounted on the end of a Hampton pin. It is possible the conductivity of the thermocouple wire itself is to blame for this difference (I recently bought a smaller one). But, at least by this kind of measurement, 8.3.1 has one of the coldest cryostreams at the ALS. ;) And, my logs say 8.3.1's cryojet was showing ~90K the whole time you were using the beamline. Anyway, the long and the short of it is: I've seen two kinds of decay rates in heavy-atom sites. It's either all over in less than 10^10 photons/um^2 (300s on 8.3.1) or things are still peachy after 10^11 photons/um^2. I have to admit I havn't varied the cryojet settings for any of my damage studies (yet). I just leave it on "maximum coldness" all the time. However, there is definitely a connection between damage rate and the crystal type. I think I told you about my two crystal forms: one dies in 300s, the other lasts for >8 hours. These were both in the same cryostream. (Wladek has seen my talk on these results... twice) Unfortunately, the "tough" crystal forms seem to be more the exception than the rule. Most heavy atoms are blown away quickly. Hope this is helpful. Say "Hi" to Brian from me! -James erica wrote: > Dear James, > I'm at Brookhaven with Brian Greensmith, and have a few questions for the > expert. I'm trying to see an anomalous signal in my Pt soak I collected at > 8.3.1 2 weeks or so ago. Does 8.3.1 have platinum mirrors? And would this > cause a signal in the fluorescence scan that wasn't due to the protein? Also, > it looks like I have a strong anomalous signal in the first 20 frames, but > that it dies off after that. Does that make any sense to you? that the > anomalous would decay faster than the crystal - and within 20 minutes? Wladek > Minor is here, and is telling me and everyone else that the problem with their > ALS data is that the cryo there isn't cold enough, and that the reaons my > anomalous decays so quickly is the fault of the cryo. Does this make any sense > to you either? > > Thanks a lot, > Erica From james holton Sent Wednesday, May 5, 2004 3:09 pm To Jamie Cate Cc Howard Padmore , John Kuriyan , Gerry McDermott , GMeigs@lbl.gov , JJPlecs@lbl.gov , jat@scripps.edu , tom@ucxray.berkeley.edu Bcc Subject Re: He and x-ray crystallography Sorry for taking so long to respond. I was out for the weekend, and then the power outage really screwed up 8.3.1's computers. It's mostly fixed now. WRT wavelength, I once measured the optimum I/sd/absorbed_dose to be between 13 and 14 keV on 8.3.1. Not quite sure why, but it could be that the detector phosphor is most sensitive out there. So, if you're damage-limited, you should be working at ~13.5keV. This will increase the exposure time, but you should get more signal/noise per unit damage. It will also compress your pattern, but you can compensate for that by increasing the detector distance. Even with the increased air path, you get a net gain in I/sd/damage. I generally run 8.3.1's cryojet at what is supposed to be 10 l/min (according to the control panel). Including the "warm stream" at 10 l/min this would be 20 l/min. Since the N2 seems to have no trouble at all equilibrating with the air outside the hutch (oxygen and humidity have no trouble getting in), I would expect that He (which moves ~50x faster than N2) would find its way out of the hutch pretty easily. Also, the only high-voltage stuff we've got is inside a 100% He atmosphere, so I don't expect arcing will be a problem. I think the biggest problem this presents is that this rate of consumption will empty a ~200 cuft tank in ~5 hours, and that might require interrupting a data set. Alastair mentioned something once about getting He bottles in "six packs", but I don't know from where. I think we can probably run the "new" cryojets with He by running a few turns of copper tube down into the lN2, and up the inside of the cold hose. We might even be able to operate in the "normal" mode this way and switch over to He as needed. I just spoke with George, and he thinks he could make the modifications in about two days. I have George tasked full time on "Cool Hand Luke" right now, but I hope we will be able to try this cryojet modification before the shutdown is over. Alternately, we could use an auxillary stream than George and I have been playing with to run gaseous He through a heat exchanger. This would involve unbolting the coldhead (burning more beam time), but might be easier than modifying the existing cryojet. We can easily hook up helium to our collimator (it was, after all, designed for it). I also have an idea for a glass capillary beam stop that we could run He through, and cover the gap between the beamstop and the sample. However, I did do an experiment comparing beamstop distances of 120mm and 10mm. There was only a ~4% difference in the I/sd at low resolution, so I decided not to worry about it very much. However, with a (non-lysozyme) crystal with higher overall background scatter, one would expect an improvement from a closer beamstop. With the detector at ~1m distance, the background from the air scatter will be maximally spread out, especially at higher angles. That is, the inverse square law applies to the noise, but not the signal. I'm wondering if the effect of He around the sample would be as noticable as it was in Bob's study. Bob was able to show a big advantage for small crystals that diffract well (~2A) at short detector distances with 8keV x-rays. Jamie has big(ish) crystals and really big unit cells and, therefore, a much larger detector distance. Anyway, as they say at NASA, one test is worth a thousand expert opinions. I think the best place to do this whould be 12.3.1, since it's easier to modify things in a big hutch. I just discussed this with JJ and he seems interested. Although 5.0.2 has a Q315 as well, it can't be moved in 2theta, and there is also the problem of having an active user program, and the chance of breaking the cryojet (for several weeks) might be an uneccesary risk. Gerry? What do you think? -James ----- Original Message ----- > Hi James, Howard, > > How are you? > > Bob Glaeser forwarded to me the reference to your Biophysical Journal > paper in 2000 on radiation damage to crystals, and the use of He in the > cold stream. As you may know (James does, I don't know if I told > Howard), we have gotten some really fantastic crystals of the ribosome > that diffract to at least 3.5 Angstroms (2 ribosomes in the A.S.U., > actually). This from crystals that have about a 50 micron > cross-section in 2 of 3 dimensions. > > I was wondering what it would take to set up an experiment with all He > in the path between the collimator and the beam stop? In the paper, I > gather you replaced the cold part of the cold stream with He, but what > about the concentric warm air part? I would really like to try such an > experiment to see if it will help us with background issues. My view > is that the beam between the collimator and beam stop is our worst > enemy, much more so than after the beam stop. Given the long exposures > we are taking, the fact that we are at ~11 keV only helps some. > > The second factor is that we are already way beyond the resolution that > we can measure effectively on 8.3.1. We really, really need the Q315 > at this point. We are using a 2-theta of a few degrees even with the > Q315. So, my dream experiment would be to set up 8.2.2 or 12.3.1 with > a full He path, to see if we can push the resolution to 3.0 Angstroms > or better, mainly by reducing the background in the images. > > We, of course, are trying to grow bigger crystals. I can say that the > ones we have cryocool well, with a mosaicity of ~0.15 degrees in the > best direction, so going bigger may not be better. > > Let me know what you think about this. I figure we have a few weeks to > gear up for a trial run! > > Best, > > Jamie > > Jamie Cate > Departments of Chemistry and MCB > University of California > Berkeley, CA 94720 > (510) 486-4033 > (510) 486-6240 > jcate@lbl.gov > From: Jamie Cate Date: Tuesday, May 4, 2004 12:07 pm Subject: Re: He and x-ray crystallography > Hi Gerry, > > Thanks for the note. James, could you tell me the L/min that > would be > coming out of the internal and coaxial flow components? How does > this > compare to the ventilation flux through the hutch? > > Jamie > > On May 4, 2004, at 1:04 PM, Gerry McDermott wrote: > > > Jamie, > > > > The only issue with He is if the concentration gets high, you it > can > > cause > > High Voltage equipment to arc. However, in reality I don't > think > > this is a > > problem. > > > > As for ventilation, we didn't make any changes to 5.0.2. I > think you > > should > > be ok in the minihutches.....although, since they are so much > smaller> volume, you might want to calculate the flow rate/hutch > volume to > > estimate > > how high the concentration might get. > > > > Gerry > > > > "Jamie Cate" wrote: > > > >> Hi Gerry, > >> > >> Thanks for your note. Here was Bob's reply to me. He seems to > think>> it is a good idea, and based on the Biophysical Journal > paper, I'm > >> gung-ho to try it. > >> > >> A technical question. Are there any worries about the amount > of He > >> put > >> out by the system? In other words, do the hutches need special > >> ventilation? > >> > >> Jamie > >> > >> Begin forwarded message: > >> > >>> From: "Robert M. Glaeser" > >>> Date: April 29, 2004 3:17:56 PM PDT > >>> To: Jamie Cate > >>> Subject: Re: He and x-ray crystallography > >>> > >>> JAMIE, > >>> > >>> REPLIES INSERTED IN CAP LETTERS BELOW. > >>> > >>> BOB > >>> > >>> > >>> > >>> Jamie Cate wrote: > >>> > >>>> Hi Bob, > >>>> > >>>> How are you? I seem to recall that you have some experience > using > >>>> He > >>>> for cryocooling crystals for x-ray diffraction experiments. > I'm not > >>>> necessarily thinking about 4+ degrees or so, but simply about > >>>> reducing background scatter. My thought is that since >95% > of the > >>>> x-rays are stopped at the beam stop, the very short distance > between>>>> the collimator and the beam stop is contributing a > significant > >>>> amount > >>>> of the background scatter. > >>> > >>> YOU ARE ABSOLUTELY RIGHT ABOUT THAT. > >>> > >>>> In our case, since we are taking multiple-minute exposures at > ALS,>>>> this is a very big problem, I think. > >>> > >>> YES, PARTICULARLY SO AS SMALLER ANGLES [AIR SCATTER IS MUCH MORE > >>> INTENSE AT SMALL ANGLES]. > >>> > >>>> > >>>> > >>>> To me, the "best" diffraction setup would be to have a helium > cone>>>> between the detector and beamstop, and all-He coming > through the > >>>> cryo-system (cold stream AND concentric warm stream). This would > >>>> reduce the background to the minimum we could practically > achieve.>>> > >>> YES, I AGREE WITH THIS. SOME KIND OF "CONE" WOULD BE IDEAL, > BUT IT IS > >>> REMARKABLY EFFECTIVE EVEN TO JUST FLOW He THROUGH THE CRYO- > SYSTEM. WE > >>> DID A QUICK-AND-DIRTY TEST BY FLOWING HELIUM GAS THROUGH THE > >>> NITROGEN-COOLED HEAT EXCHANGER. > >>> > >>> FIGURE 5 OF > >>> Glaeser et al. (2000) Biophysical Journal 78:3178-3185 > >>> DOCUMENTS THE IMPROVEMENT ON BING JAP'S ROTATING ANODE MACHINE IN > >>> DONNER. GERRY MCDERMOT HELPED US TO SET UP MEASUREMENTS ON BL5 > (THE>>> CENTRAL HUTCH), ONCE WHEN WE HAD TINY MICROCRYSTALS, BUT > WE DID NOT > >>> CONTINUE BECAUSE THE CRYSTALS WERE SPOTTY. > >>> > >>>> > >>>> > >>>> Let me know what you think, and if you have some information > on this > >>>> kind of approach. > >>>> > >>>> Thanks! > >>>> > >>>> Jamie > >>>> > >>>> Jamie Cate > >>>> Departments of Chemistry and MCB > >>>> University of California > >>>> Berkeley, CA 94720 > >>>> (510) 486-4033 > >>>> (510) 486-6240 > >>>> jcate@lbl.gov > >>>> > >>> > >>> > >>> > >> Jamie Cate > >> Departments of Chemistry and MCB > >> University of California > >> Berkeley, CA 94720 > >> (510) 486-4033 > >> (510) 486-6240 > >> jcate@lbl.gov > >> > > > > > Jamie Cate > Departments of Chemistry and MCB > University of California > Berkeley, CA 94720 > (510) 486-4033 > (510) 486-6240 > jcate@lbl.gov > > From JMHolton@lbl.gov Sent Wednesday, June 16, 2004 2:15 pm To Yu Chen Cc GMeigs@lbl.gov , JTanamachi@lbl.gov Bcc Subject Re: question about data collection The data collection might be pausing because the flux at 14keV is lower than expected, and the control system thinks the beam dumped. If you push "optimize beam", that should bring it back up. I'll look into this to make sure it works before Friday. you can use 2theta values of up to 41 degrees right now. Probably not a good idea to bring it so high that you can't see the beam center on the detector surface. You'll have to do a little math and/or trial and error to figure this out. I'd stick to 10 or 20 degrees. Also, do NOT do a fluorescence scan with the detector at non-zero twotheta since there is a collision hazard with the fluorescence detector. Other than that, the image headers should contain everything you need to process data at 2theta angles. -James ----- Original Message ----- From: Yu Chen Date: Tuesday, June 15, 2004 11:47 am Subject: question about data collection > Hi James, > > I am going to beamline 8.3.1 this Friday and I have a few > questions on > collecting ultra-high res data. My crystals diffracted to below 1A > at your > beamline before and the spots were pretty strong even at the edge > of the > image plate. But when I tried to lower the wavelength to below > 14000ev,the shutter could not open. Could you tell me whether > there is anything I > should pay attention to when I collect high-res data at beamline > 8.3.1?Also, I am fairly inexperienced when it comes to collecting > such high res > data. It will be great if you could give me some suggestions, like > what2theta values people usually try, etc. > > Thanks a lot! > > -Yu > > From JMHolton@lbl.gov Sent Saturday, June 19, 2004 8:58 pm To beamline-support@lbl.gov Cc Bcc Subject Saturday Night 2 8.2.2 Complaint: last time, all of my low-res data sets had really high R-merges. What can I do to prevent this? Description: Cynthia is trying to limit the radiation damage to her crystals. She is working at 14-16keV. It is difficult to deconvolute from the dataset taken, but all of Cynthia's short-exposure passes (0.5s) on the 13th appear to have been unusually noisy. In the past, I performed a test on 8.3.1 and found that exposure times (at 11 keV) as short as 0.1s had no negative impact on data quality. We did notice the striations on the bottom middle module that Borders(TM) already described. The images from the 13th also have these striations. However, re-processing with the affected module blacked out of HKL2K did not significantly help the Rmerge. (12% -> 11% instead of 4% for exposures > 1s) Actions: After some screwing around, we finally took a pair of datasets to test the performance of short vs long exposures: set exposure Hdiv Rmerge Rmerge(4A) 1 0.3 3.0 13.8% 9.0% 2 3.0 0.3 8.8% 5.1% at 200mm detector distance, the impact of Hdiv on spot shape should be minimal. I checked for discrepancies in the spindle/shutter control I5xx variables between the tree and the pmac itself, but found none. Outcome: It appears that exposure times less than 1s on 8.2.2 are inadvisable at this time. Someone should collect test data to evaluate the performance of short exposures on 8.2.2. It is possible that the combination of short exposures an high energy (~15keV) could have something to do with it. I.E. there may be significant vibrational components to the 15 keV beam at ~1 Hz. Someone will have to do a power spectrum to find out. The bottom middle module of 8.2.2's detector should probably have its power supply replaced. Another complaint: Cynthia could use a way to record the flux of the beamline so she can try to control for flux in her damage assesments. I know of no easy way to do this on 8.2.x right now. I used to have "flux.com" working on 8.2.x, but none of my stuff works on 8.2.x anymore. Also, The diode in 8.2.2 has serious light leaks, which makes it difficult to figure out what the actual flux is. Painting the foil seams with some of my black carbon paint is an easy way to eliminate light leaks. I'd be happy to show anyone who is interested. -James From james holton Sent Friday, August 6, 2004 2:30 pm To Chuck Farah Cc GMeigs@lbl.gov , JTanamachi@lbl.gov Bcc Subject Re: Chuck Farah Each Wedger run will independently try to figure out the space group, and they do sometimes arrive at different conclusions. This will not adversely impact further processing as long as all the wedges are on the same point group (which they are). Scaling P2 is absolutely identical to scaling P21, with the exception that a few spots are supposed to be zero intensity. You can process your whole dataset in P21 and calculate maps in P2 and never know the difference. In general, if I've learned anything about space group selection it is that absences LIE! Systematic absence interpretation has led as many people astray as they have helped. Remember, you throw out more observations due to scaling rejects than you have observations of systematically absent spots. Why do we base such a critical decision in such little data? The first time the difference between P2 and P21 becomes important is phasing. Even then, it can be difficult to tell. The first time in any project that you are absolutely sure of your space group is when you see an interpretable electron density map. Elves will consider P2 and P21 to be equivalent until the very end. By, default, solve and epmr jobs will be run in both space groups, regardless of what the Wedger jobs found. You can force Wedger to use a particular space group by putting it on the command line when you launch Wedger, or when you launch Elves or Processer. This will disable the automatic space group search. If Wedger arrives at different point groups for different wedges, then something is probably wrong with the processing (radiation damage or wrong camera parameters, etc). So the wrong point group should draw your attention to "problem" wedges that might need manual intervention. Hope this is helpful, -James Chuck Farah wrote: > Hi James, How are you. Thanks for the info. We have been looking at the processing output of the datasets for the XAC2396 protein produced by Wedger Elves and found a few discrepancies that you may be interested in and about which I would like your opinion: Maybe you remember that we took two datasets initially:1) br2396 dataset which was collected at two wavelengths (peak and low) with two 100 image wedges at each2) br2396_hires dataset was collected at one wavelength (peak) with two 100 image wedges When we did a mtzdump on the merged.mtz file in each of the wegdes we found that: 1) in the br2396 dataset the two low energy wedges were processed in P21 while the peak energy wedges were processed in P21 for wedge 1 but in P2 for wedge 2 2) in the br2396_hires dataset the the peak energy wedge was processed in P21 for wedge 1 but in P2 for wedge 2 Is this due to a glitch in the script or did we do something wrong during the processing? I expect that this would cause problems in te subsequent data anaylsis. I would love to have your opinion in this regard and any advice. All the best, Chuck São Paulo, Brazil > > ----- Original Message ----- > From: james holton > To: Chuck Farah > Sent: Thursday, August 05, 2004 3:39 PM > Subject: Re: Chuck Farah > > > > The "def.site" file for hkl2k here is: > > http://bl831.als.lbl.gov/def.site > > We have an ADSC Quantum 210 detector. All data are pre-transformed and binned. The goniostat is in the standard "normal beam" geometry and rotates in the same "direction" as most spindles in the US. > > Hope this helps! > > -James > > > Chuck Farah wrote: > >> Hi James, How are you? We are trying to process the data that we collected last week using HKL2000 and we need a few technical details regarding the detector and the goniostat that was used. Specifically, we need to know - What specific type of ADSC detector? Q210 or Q315 or Q4/Q4R and transformed data or raw data- Which goniostat type? Thanks! Chuck FarahUniversidade de São PauloBrazil > From james holton Sent Sunday, December 19, 2004 1:27 pm To BSSchuwirth@lbl.gov Cc KAFrankel@lbl.gov , GMeigs@lbl.gov Bcc Subject Re: strategy The "strategy.com" file should look just like the one that Wedger Elves produces. This will be appropriate for the particular detector size and camera parameters you are using. To use "strategy.com" with "incremental_strategy.com", you need to make sure the "strategy" lines in the script look like this: strategy start 0 end 180 speedup 1 go stats end The "speedup 1" thing is most important because it prevents mosflm from monkeying with the unit cell parameters. -James BSSchuwirth@lbl.gov wrote: > Hi James, > > I'm not sure if you will get this message before you leave for hawaii. > Ken mentioned to me that we need to have a new "strategy.com" file in > case we are changing the 2theta. Could you please confirm that? Maria > and I never did it but the program still seemed to work. > Thanks + have fun on hawaii > > Barbara From james holton Sent Tuesday, January 11, 2005 12:53 pm To Dan Minor Cc GMeigs@lbl.gov Bcc Subject Re: Zinc MAD I'm afraid the scans you have look pretty weak. (I'm assuming test_Zn_041.scan was the zinc reference) Flux optimization will not have any impact on the fluorescence scan. In fact, the scan program always attenuates the beam by a factor of 100-1000 or so before doing a scan. Otherwise, the raw counts would saturate the fluorescence detector. What you're seeing is a healthy number of raw counts (30,000-50,000), but the incremental jump over the zinc edge is small. This means that you have a lot more non-zinc atoms in the beam than you would like. Probably around 1 mM Zn in the sample, which is about our detection limit. You can figure out how much zinc you should have with "cellvolume.com" on bl831: CELL 90.198 98.088 98.243 90.281 110.691 115.006 SYMM P1 MASS 75479.8 CHAINS 4 VM 2.4 SOLC 48.35% CONC 9.2 mM if you really do expect 9.2 mM Zn, then your scan should look a lot stronger than it does. Might be an interesting excercise for a student to prepare a Zn solution that has the concentration of Zn that you would expect to find in your crystals and check that fluor scan when you get here. ;) It is also a possibility that you have no zinc at all and you're seeing residual signal from the ZnS in the beamstop. I don't THINK this is a problem, but I don't have a "no-sample" zinc scan on hand to prove it. I'll run downstairs and do that now. The door problem (from what George tells me) turned out to be an old ghost from the past. It's been at least a year or two since we had the Lantronix ethernet-to-serial interface on the motor controllers flake out on us, but that is what happened. I think it has something to do with the connector (custom made at LBL), but it happens so rarely, that I havn't been able to figure out what the real cause is, and we can't "make" this problem happen either. Makes things hard to debug. Dan Minor wrote: > Hi James, > > We had a good run on Saturday night. We are collecting data on a new > complex. We know there should be four zinc atoms per complex. We > were checking by XAFS on Saturday. There was an edge at the right > position, but it was a small signal. We did a control with the ZnS > sample (which gave a huge signal). I'm hoping that the small size of > the signal from our crystals was because I didn't do any flux > optimization at the Zn wavelength (9663 eV or so). We were doing all > of our collection at 11111 eV. That seems far enough away that the > flux could be the answer. I wanted to check with you before we come > up tomorrow night to see if there are any other things we should know > about 8.3.1 and the zinc edge. We plan to go for a Zn MAD set. > > Hope the door problem was easy to solve. I felt bad about bothering George. > > Best, > Dan > -- > > ________________________________________________________________________________ > > Dan Minor, Ph.D. > Assistant Professor > Departments of Biochemistry and Biophysics, > and Cellular and Molecular Pharmacology > Investigator, Cardiovascular Research Institute > University of California, San Francisco > 513 Parnassus Ave. Room HSE1308 > Box 0130 > San Francisco, CA 94143-0130 > Email: minor@itsa.ucsf.edu > Phone: 415-514-2551 > Fax: 415-514-2550 > Web: http://www.cvri.ucsf.edu/~dminor From james holton Sent Wednesday, January 12, 2005 2:34 pm To Dan Minor Cc GMeigs@lbl.gov Bcc Subject Re: Zinc MAD Yesterday, I checked a "no sample" scan of Zn, Ga and Fe. The Ga scan is really flat, but the Zn and Fe scans had little wiggles in the middle. Probably more from the steel in the beam pipe, etc than from the beamstop. Your scans, althrough weak, still look more "edge-like" than these "no sample" scans. -James Dan Minor wrote: > James, > > Thanks. Puzzling. We actually have eight zincs in the unit cell (so > our puny signal really is only at 5% of expected). We haven't > included the zinc in our preps, but will on the next one. Our > crystals are real rocks. I'm surprised that things are so stable > without the structural zinc.... > > We'll scan an 18 mM Zn solution to get a better metric tomorrow. > > I thought the door problem might the one you mentioned. I remember > you mentioning this problem before. Sounds like a hard one to hunt. > > Dan > > -- > > ________________________________________________________________________________ > > Dan Minor, Ph.D. > Assistant Professor > Departments of Biochemistry and Biophysics, > and Cellular and Molecular Pharmacology > Investigator, Cardiovascular Research Institute > University of California, San Francisco > 513 Parnassus Ave. Room HSE1308 > Box 0130 > San Francisco, CA 94143-0130 > Email: minor@itsa.ucsf.edu > Phone: 415-514-2551 > Fax: 415-514-2550 > Web: http://www.cvri.ucsf.edu/~dminor From GMeigs@lbl.gov Sent Tuesday, February 22, 2005 5:19 pm To GMeigs@lbl.gov Cc Bcc Subject finished_run to complete a DVD write without typing "finished" do this: touch /data/mcfuser/finished_run this will create a file called "finished_run" which backup_rimage.com looks for every 10 seconds; when it finds the file it flushs the data to the DVD and deletes the "finished_run" file. From james holton Sent Wednesday, June 15, 2005 11:40 am To Jeremy Wilbur Cc GMeigs@lbl.gov , JTanamachi@lbl.gov Bcc Subject Re: phasing advice First thing you want to do is look at your I/sigma ratios for the datasets you've got so far. If you're doing MAD or SAD then you're going to need: I/sigma > 1.3*sqrt(kDa/sites)/f" This "threshold" is where the signal-to-noise ratio of the anomalous differences becomes one. Suprizingly, you can solve MAD/SAD structures even with a signal-to-noise rato that is that bad, but you can't go much lower than that. The unfurtunate thing about this, however, is that when DANO/SIGNADO ~ 1 there is no way to tell IF you have an anomalous signal at all. If you plug S SAD into this equation you can probably see why it almost never works. Then there's the "damage limit". Native crystals don't absorb as much energy as heavy-atom soaked ones, and the concentration of heavy atom can very strongly effect crystal lifetime. Unfortunately, there are no good data on what you can and can't get away with. Empirically, for the crystals I am using to study this, you seem to always be safe with < 300s or so of exposure in 8.3.1. you might be able to get away with as much as 1000 or even 3000s in favorable cases. I am developing a procedure to try and preemtively measure how much energy a crystal is absorbing and figure that into a data colleciton strategy, but I havn't got this working yet. Rule of thumb is 300s. You will probably be left with the situation where 300s will not give you the I/sigma you need. This is why solving structures is so hard. In general, I would recommend the philosophy of one crystal at a time, and see how often you can get yourselves over here. Each of your crystals will last not more than ~1 hour in the full x-ray beam, so it shouldn't be hard to figure out what your data collection time is going to be. (Note, reducing the divergence will proportionally increase the expected sample lifetime). We have had one success (Clare Peters-Libeu) using Pt for RIPAS phasing, where you use an intentionally damaged dataset as a "native" for phasing. Others around the world have reported a few sucesses with RIPAS using Iodine. It's a strong signal when it works, but you need to have the situation where your heavy metal decays a lot faster than anything else. Otherwise, non-isomorphism kills you. Much like it does in MIR. Anyway, I hope this is helpful in some way. Let me know if you have any more questions. Jeremy Wilbur wrote: > Hi James, > I'm an 8.3.1 user from the Fletterick lab. Peter Hwang and I are > trying to figure the best approach for the situation I'll describe > below. We thought we'd see if you had any advice or ideas. > > The situation is that we have collected three data sets from three > crystals each to about 2.5A resolution. The data look good so far. > Unfortunately, I haven't been able to reproduce these crystals so all > we have are native crystals. We are think this is because of some > degradation occurred that allowed the crystal formation. We're working > on figuring this all out and maybe making new constructs. In the mean > time we're in a bit of a race with another group to solve this > structure. We have about a dozen more crystals in the drops. We need > to collect data to phase these things. Our thoughts were to try the > standard soaks of heavy metals but we don't want to waste to many of > our remaining crystals trying things out. We also thought about > bromide and iodide soaks. Finally we thought about sulfur SAD. > Incidentally elves, thought the space group was P222 or P22121. > > We have little no experience with the halide and sulfur methods and we > were wondering if you have such experience or know if people have tried > this on 8.3.1. > Also, do you think these are worth while to try out or not? > > Anyway, any thoughts, pointers, ideas, on the matter would be much > appreciated. We're in the planning stage right now and have beam time > again Sun 6/26 4pm. > > Thanks, > Jeremy From James Holton Sent Monday, October 17, 2005 6:33 pm To Betty Shen Cc Bcc Subject Re: space group ambiguity Betty Shen wrote: > > On Oct 17, 2005, at 12:41 PM, James Holton wrote: > >> >> SAD damages your crystals just as fast as MAD. In fact, SAD can be >> even worse than MAD because you are sitting at a wavelength where >> your crystal is absorbing the maximum amount of energy (this is only >> significant with Se-rich xtals). The only advantage with SAD is that >> you collect fewer images, so there is less total dose than a MAD >> experiment using the same exposure time. However, doing a >> 2-wavelength experiment with half the exposure is better than doing a >> 1-wavelength experiment in the controlled experiments I have done >> with my model systems. >> > True., and I did collect a three wavelength MAD data set on a separate > crystal. But perhaps I made the mistake of collecting all three > wavelength simultaneously. > Should I have collected them consecutively? NO. Measuring f' by subtracting decayed data from undecayed data in MAD analysis is a really bad idea. The differences due to radiation-induced changes in the structure are seldom useful in phasing (although Ravelli et al. had some success by specifically radiolyzing an Iodine and using those differnces for phasing). If you're going to do MAD, then you should collect all the wavelengths simultaneously. That is, the two spots you are going to subtract to get anomalous and dispersive differences are best collected close together in time (for a large number of reasons, including changing beam intensity). > The reason I rely more on SAD than MAD is because all MAD data set I > collected so far have very poor correlations. In your case (strong expected signal) this is probably due to damage. Use shorter exposures. >> (what was your Se content, and did you have any atoms heavier than >> oxygen in your cryo?, if so, how much?). > > > I have 6 SeMet for a protein/DNA complex with approximate Mw of 29Kd > with strong Se signal in the fluorescent scan. > (I did convert all Se to the oxidized form to sharpen the signal. It > worked beautifully in previous experiments). > >> >> What was your Se content and what is your I/sd ratio? > > I/sd at 2.5A is 85/2.1 = 40 > and 1.3*(30kd/6)/f'' = 24 > >> >> If this isn't cubic then it's really really close. > > The reason I think that it isn't cubic is because the unit cell > parameters are too small to accommodate even one complex/ASU. If I had a nickle for every time I heard those words..... Try cubic. Get it over with. I'm sorry to tell you that you probably don't have a crystal of your complex. If the data merge well in a cubic space group, then the crystals are cubic. End of story. > I choose F222 because it has the next highest symmetry that can > accommodate one monomer/asu. The self-patterson has a two-fold NCS. > I thought that was because the molecule itself is active as dimer. Be VERY VERY CAREFUL with NCS if you are dropping crystal symmetry. One of the most evil things you can do as a crystallographer is throw out a crystallographic symmetry axis, pick a free-R set, and then "apply NCS" across what should be crystallographic symmetry. If you do this, then every HKL in your "free" set will actually have an "NCS symmetry mate" in the working set. Your Free-R in this case will drop like a rock, and be absolutely invalid. You could build you sequence in backwards and never know it. >> >> Since you collected 180 degrees, I recommend you cut back on your >> frames. Use only the earliest data possible that still gives you a >> complete data set. Then check and see if you have an anomalous signal. > > I will give it a try. > Good luck, -James Holton MAD Scientist