Automated data collection Peter Zwart Berkeley Center for

Automated data collection Peter Zwart Berkeley Center for Structural Biology Physical Biosciences Division, LBL

Automated data collection • Screening of crystals is a typical task that can be (semi) automated, especially when a robot is used. • The benefits of automated screening primarily lie in – – Reduction of human errors in repetitive tasks – • Efficiency in screening Tracking the results DCS already had/has quite a number of sophisticated tools that allow HTP screening and collection – • It however lacks fully automated analyses of images and has other drawbacks The Web. Ice server allows an automated characterization of X-ray images via an analyses done with the Labelit suite (indexing, automated image quality, point group assignment)

Automated data collection • As part of a major BOS upgrade, Nick Sauter, Yun Zhou, John Taylor, Paul Adams and myself have been working on implementing automated screening and data collection facilities in BOS. • Key ingredients: – A data base that holds user provided sample information (which crystals in which puck in which position) – An external server providing data analyses • Indexing, diffraction quality, resolution estimates, data collection strategy – Automated crystal centering software (XREC) – BOS – Gyorgy Snell & Josh Stillwell

Screening & Data collection Sample information (Xtal ID, Puck name, Position in Puck) Samples in dewar Generate new run list Samples in run list Mount next pin Analyze data Web. Ice, manual (auto) center All done? Take snapshots Collect / Continue Unmount pin ?

Automated crystal centering • XREC is used for automated crystal centering • It works well on 5. 0. 1 • 5. 0. 3 has some lighting issues that need to be addressed • Auto centering is a two stage procedure – Finding and centering the loop in low mag – Finding and centering the crystal in the loop in high mag • An iterative approach is used and is relatively robust • Automated centering usually takes less than 90 seconds – Manual centering is still faster though

The new interface • Center/Robot Tab • Maps puck name to robot dewar position • Familiar centering interface • Log window: Shows all what happens – A ‘light’ version of the logfile (with mounting, unmounting etc info) will be available in the near future • A scoring tab allows efficient manual classification of diffraction quality

The new interface • Tasq. Queue Tab • The interface between the database and the BOS user • Search for puck names and add the crystals / pucks to the Queue • Decide in which SIL (Sample Information List) you want Web. Ice to store the results of the analyses • Start the data collection by pressing START

The new interface • After starting the Queue, crystals will be mounted • User will center crystal • or XREC will run • 0 and 90 snaps are taken (and xtal jpegs are saved) • Images fed to Web. Ice for analyses • Current crystal is unmounted, next crystal is mounted • Etc etc • The Web. Ice tab in BOS shows the results of auto indexing, resolution estimates and gives an overall score to the two images • The system works reliable – I have used it to screen about 6 pucks of real crystals myself • From my sofa

Web. Ice • Web. Ice allows you to view – – • the snapshots the results of the analyses From any where as long as you have an account – Your LDAP name and Password • Data are grouped in SIL’s

Communicating with the data base • How to populate the data base with your data? – – Spreadsheet upload (under development) – • Manual entry Other External application Extracting information from the data base – – • External application Spreadsheet download Issues – Spread sheet uniformity / standards

Conclusions & Acknowledgements • Automation efforts in BOS are in a stage close to beta testing • Acknowledgements – John Taylor – Yun Zhou – Nick Sauter – Paul Adams – Ana Gonzalez – Boom Moorhouse – Gyorgy Snell – Josh Stillwell