=================================================================== COMMENTS FROM THE ACA MEETING IN ST PAUL MINNESOTA 22ND - 27TH JULY 2000 =================================================================== This is rather a long document, so to help you it is split into two parts: I : Exhibition II: Sessions The "Exhibtion" part covers details from organising and running the stand in the commercial exhibit, plus details of feedback from visitors. The "Sessions" part covers what we learnt from the scientific sessions - apologies if it is too long/detailed. =================================================================== I. EXHBITION ============ 1. INTRODUCTION --------------- CCP4 staff Alun Ashton and Peter Briggs attended the ACA 2000 meeting in St Paul Minnesota, and manned a booth in the exhibition. Also present were Harry Powell (MOSFLM author) and Katherine McAuley (CCP4 user). 2. EXHIBITION STAND ------------------- Background: ----------- The ACA committee kindly offered us a complementary exhibition booth after we approached them asking for space to display our poster. We were required to pay for additional furnishings e.g. tables, poster boards etc ourselves. The arrangement used for the stand consisted of: 10'x8' poster board, 4 chairs and 2 small tables. We used the "pannelled" CCP4 poster (originally made for the IUCr 1999 meeting) with the A0 panel updated for CCP4 v4.0, plus additional small panels for: MOSFLM, CCP4nt and Data Harvesting. The CCP4 laptop was used for demonstrations of CCP4 on one table, Harry Powell used his laptop on the other for demos of MOSFLM. What worked? ============ i. Harry/MOSFLM was very popular - lots of people are interested in trying MOSFLM. ii. A4 printouts of the CCP4 v4.0 poster - which highlighted new and updated programs and features of the latest release - and of the MOSFLM poster were very popular. All these were taken by the end of the 4 days. iii. Manuals and leaflets were all gone by the end of the exhibition - we took approx. 100 manuals and could easily have shifted more. iv. Having a CCP4 user on the stand was handy, and contributed knowledge that the programmers lacked (Harry's background knowledge as a former crystallographer also made a significant contribution in this regard). v. Having a laptop and working demos was very useful. What didn't work? ================= i. The timing of the latest Refmac release etc didn't work in our favour - we were unprepared and were unable to answer many of the questions people asked about it. ii. Small posters on CCP4nt and Data Harvesting didn't seem to attract much attention (maybe because we were stood in front of them? :) What could be done in future? ============================= i. A2 poster on Refmac - and turn this into a leaflet as with MOSFLM and the v4.0 poster. Maybe more of these hand-out style leaflets for other programs too? Or even at a higher level, e.g. MR using CCP4, or the flow chart/roadmaps? ii. More communication with developers - they could give us more information/demos etc and emphasise the points that we should be passing on to the users. iii. We tried a "comments book" in Glasgow, with limited success but I think we should try this again - get people to write likes/ dislikes etc (anonymously if they want, or leave contact e-mail if they would like to follow it up). Alternatively we could try a small (A5-size or smaller) questionnaire. 3. (DIS-)ORGANISATION --------------------- When we arrived there was no equipment i.e. chairs, poster boards etc - the order was delayed by Rutherford and had never reached the exhibition contractors. The delay might have been avoided if we had sent the order to Rutherford even just a few days earlier, and we might have had an idea that there was a problem if we had followed it up before leaving for the States. Advice for avoiding this in future: i. (Pretty obvious) get everything sent away as soon as possible. ii. Follow up the orders once they have been sent to Rutherford. Delivery of the posters, manuals etc was not a problem - however getting the posters back home was more of a problem and we ended up carrying them back on the plane. If we do this again in future then we need proper packaging etc for this. 4. USER FEEDBACK ---------------- Prior to the meeting Alun and David sent out letters to the American companies advertising our presence at the conference (we also sent a message to ccp4bb). Obviously it is difficult to assess the impact this had, but our stand did recieve a large number of commercial visitors. Alun remembers that the ACA also had a significant number of Japanese industrials, so next time we should also send letters to them. As well as the commercial users, we got academic users visiting the stand. Some people had specific questions about the software, others wanted demos of the gui or (more usually) of MOSFLM. A few people also came over just to say "hi!" General comments/questions: --------------------------- * A lot of people are using the programs and the gui successfully and are very happy with it. A few people (e.g. Mark Elsinger) are using the latest beta gui and had positive comments. * Many people had questions about REFMAC5 - unfortunately we were unable to provide a lot of info. * People were still asking for demos of the gui, increasingly for demos of particular tasks (e.g. amore/molrep, patterson) [Many people reported that they were using the gui successfully and the general impression is that we have already won many converts to the ccp4i cause. The downside is that I think there are many of the extra features (e.g. "project management tools") which people are not aware of and yet they think they "know it all" as far as the gui is concerned - so they're no longer asking.] * There were a few complaints about the documentation but the majority of the comments were very positive (some people claiming that they learnt everything from it - an obvious fabrication! :) - so on the whole we must be doing okay in this department. * Some Americans seem to use CCP4 for small tasks but they don't really like it - we had a few complaints, particularly about MTZ format and the file labels, and (percieved) difficulty in looking at the file contents and importing/exporting (a lot of people use CNS). This may be a minority! When questioned they often didn't seem to know about mtzdump, so this is possibly an "education" issue (maybe we need a "beginners guide to CCP4" - "CCP4 for morons", anyone?). We tried to promote the gui in this context too, with a positive response. Specific comments/questions: ---------------------------- * The fsplit source should be distributed with the suite. * Complaint that the new XDLview breaks on alphas. * References to the PDB Brookhaven (in the manual) are out of date. * The "linux" and "linuxppc" options in configure should be the same(?) * Request for MAC OS-10 support. * Talady Bhat - talked about the CCPn initiative and suggested that a representative of CCP4 should go to the next meeting at NIST(?) =================================================================== II. SESSIONS ============ General impressions: obviously this conference was not just PX. The general themes/areas of interest seemed to be related to "structural genomics" and high-throughput structure determination. Almost all the emphasis was on the crystallisation and data collection side, practically nothing on the data processing/structure solution side. High-throughput crystallation ----------------------------- The crystallation step is the bottle neck in the structure determination process. High throughput will be needed for structural genomics - the aim is to do as many successful crystallation experiments (i.e. produce well-diffracting crystals) as quickly as possible. Automation here means: robotics (to automate setting up experiments), methods for facilitating or even automating the analysis of the outcomes (i.e. did we get crystals? were they any good?), and collecting information on large numbers of experiments in order to identify patterns of similar properties and crystallisation outcomes ("data mining algorithms" and "case-based reasoning"). Brent Segelke (Lawrence Livermore) presented results which suggest that "random screening" is the most efficient way of sampling the different crystallisation parameters ("efficient" in this case means "least number of experiments performed on average in order to get decent crystals", compared with the grid method, where a tray is set up with progressive variations in conditions). Their program CRYSTOOL for automating the design of random crystallisation screening, combined with commercially available dispensing robots and automated analysis software (e.g. CRYSTALSCORE, also commercially available from Diversified Science Inc.) He predicted that dispensing will be the rate-limiting step. Lance Stewart (Emerald Biostructures) presented CrystalMonitor, which is a relational database application for HPT crystallisation. The concept is similar to the previous idea, the database records information about the conditions and results of each crystallisation experiment. Speech recognition software is used to facilitate rapid annotation of large numbers of experiments, there is also high-resolution image capture and analysis for automatic detection of crystal growth. The database can be queried to try and determine factors leading to successful crystallisations. The database is extensible and he outlined plans to use it to capture additional information from other parts of the structure determination process (e.g. CCP4 Data Harvesting files). There were also plans for integration with other software, robots etc. High-throughput data collection / remote data collection / service crystallography ------------------------------------------------------------------------- Bob Sweet talked about "courier-base data collection" at the NSLS - this is a shift away from the idea of the synchrotron as a "user-facility" into a "remote data collection facility". This done with the aid of "collaboratory software" which displays the current state of the experiment over the web, and allows remote control of the beamline and data collection apparatus. Web-cams provide "telepresence" for remote users and audio communication (via an internet "walkie-talkie") is also available. Collaboratory software allows new working methods up e.g. professors at home advising students on the beamline, co-workers monitoring data collection and reduction remotely, beamline staff supporting users without having to be at the lab. One future possibility is for complete data collection from a remote site with a technician mounting specimens at the synchrotron ("courier-based data collection"). Waldek Minor talked about data collection and processing via the internet, which seemed similar to the picture painted by Bob (i.e. remote interaction with the experimental setup allows the synchrotron experiment to be performed from home) with the extra requirement that sample mounting and alignment should also be automated (rather than being performed by a human technician). He also stressed that all the computational tasks should be integrated so that the user doesn't spend additional time transferring between different packages for data collection and data processing. Liz MacLean talked about DARTS in the context of service crystallography at synchrotrons. Experimental samples are sent by users of the facility, synchrotron staff then perform the requested experiment and write a report, send back data etc. Advantages for users is that the experiments are performed by expert staff, also there is a rapid turnaround time since beamtime is pre-booked for DARTS work. Unlike the remote data collection/ processing there is no attempt to give the user access to the experiment, only to the result. Other interesting snippets: --------------------------- Using single anomalous scattering (SAS) from sulfur (Gary Newton): the idea is to use the anomalous signal from sulfur atoms in the crystal to determine the sulfur positions and hence generate phases to give an interpretable map. This would be useful in cases where it is not possible to create useable heavy atom derivatives. Structure determination from sulfur SAS has been successful using synchrotron data but wasn't demonstrated successfully when using data collected from a home source (the ultimate aim is to use a home source). Problems: at wavelengths available from home sources and synchrotrons the anomalous signal from sulfur is very weak compared to that from heavy atoms like Fe, Se etc. A MAD experiment is not possible because the absorption edge of sulfur is 5A (synchrotron can manage 2A). High symmetry systems are better because of the high redundancy. You need some way of breaking the phase ambiguity (they used a program called ISIR/ISAS, which uses solvent flattening to resolve the ambiguity - but OASIS could also be used?). Protein crystal engineering (Jefferey Bell): tried to improve crystallisation probability by changing amino acids in the protein sequence, by using rules derived from "knowledge based potentials" (a measure of whether an amino acid is involved in crystal contacts more or less frequently than would be predicted from random pairings of proteins). Ultimate aim is to engineer proteins to crystallise in a desired lattice/packing orientation. Limited success so far. =========================================================================== Peter Briggs Alun Ashton