The production of high quality, individually characterized

Transcription

1 Managing the Workflow of a High-Throughput Organic Synthesis Laboratory: A Marriage of Automation and Information Management Technologies By Burton A. Goodman, Ph.D. Molecumetics th Ave. N.E. Bellevue, WA bgoodman@molecumetics.com ABSTRACT The production of high quality, individually characterized libraries of compounds requires a welldeveloped strategy, along with robust, userfriendly data and automation systems to support this strategy. We developed a workflow process to improve the efficiency of producing greater than 50,000 individual compounds per year. Our workflow process eliminates bottlenecks in traditional procedures and thus ensures a steady, predictable compound production rate. Initially, virtual library creation allowed for the determination of the physio-chemical properties of libraries as well as the ability to calculate the needed amounts of reagents and solvents for a production campaign. At this point, we registered all samples into our library database. Production of libraries began with loading the appropriate linkers onto resins, followed by complete characterization of the resin-linker intermediate. The resins were then loaded into IRORI MiniKans and sorted using the IRORI Autosort 10Kx sorting workstation. Chemistry was then performed using standard laboratory glassware. Additional sorting and chemistry steps were then performed until the compound was ready for cleavage off of the resin. The MiniKans were then sorted into Bohdan MiniBlocks and treated with the appropriate cleavage cocktail. Collection into 48-position racks was followed by removal of cleavage solution through vacuum centrifugation. The concentrate was then dissolved in a solvent mixture that allowed for standard liquid handling automation to create 96 well plates for analysis by high throughput flow inject NMR and LC/MS. Compounds> 80% pure by LC/MS and presenting a >80% overall yield, as determined by NMR, were automatically flagged acceptable in the database. Compounds that failed to pass these analytical criteria were flagged fail in the database and were examined for trends that could be corrected in future library production. Finally, acceptable compounds were reformatted from the original source plates into master plates. INTRODUCTION In August, 1998, Molecumetics established the High Throughput Organic Synthesis (HTOS) group with the first year goal of producing >50,000 high quality, individually characterized compounds. The challenges in this goal included specifying, installing, and bringing into production the necessary hardware and software for these tasks as rapidly as possible with limited resources. The tight time constraints of our goals necessitated that we buy a complete system instead of taking the time to design and build a new combinatorial synthesis platform. Fortunately, we could leverage our in-house experience with data management systems including ISIS and Oracle. Chemistry was similarly leveraged by relying on Molecumetics extensive experience in solid phase synthesis. Our lack of existing legacy software systems was a distinct advantage in accomplishing our goals. Our strategy was to simultaneously implement software and hardware systems, thus ensuring a marriage of automation and information management technologies. Of critical importance was choosing the data management software for library generation and production. After evaluating several commercially available software products, we standardized on the Afferent, software, which is a reactionbased combi-chem software management system. This software allowed for easy import of building blocks from our existing ISIS/Oracle inventory system, enumeration of the combi-chem products for both virtual library analysis, and library production, and as a further benefit, seamless interaction with library production automation software. In choosing the automation to support library production using solid phase as the primary synthesis technique, we evaluated two possible approaches: a high capacity parallel synthesis approach, or a split and mix approach. We tested and decided to implement the IRORI split and mix approach that utilizes minireactors containing an RF tag and solid phase synthesis resin. This approach allowed for maximizing the benefits of split and mix synthesis while allowing for production of single compounds per well. Individual compounds were obtained following a final array of the micro-reactors in a spatially addressable MiniBlock as the

2 Figure 1. Data Management Strategy Figure 2. Production Strategy cleavage block. As an added benefit, the software for managing the synthesis had built-in integration with our chosen chemistry enumeration software, strengthening the interaction of the hardware and software systems. The downstream compound handling strategy allowed for use of commercially available liquid handing automation and drying systems. Finally, we specified as our goal the analysis of every sample using high throughput LC/MS and NMR. The choice of these analytical systems was influenced not only by their ability to provide the desired analytical information but also with the ability of the data systems of these instruments to interact with our data management strategy. EXPERIMENTAL We first developed an information management strategy as outlined above in figure 1. In this strategy, building blocks registered in our in-house inventory system (a set of reagents chosen for their chemical diversity, availability, cost, and classified by functional group) were transferred to Afferent ures in the reaction based Afferent library building module of the software. The validation of the synthetic pathways leading to these templates is recorded in our corporate chemical database. Using Afferent, virtual libraries are enumerated using large sets of building blocks. We typically generate virtual libraries 3-4 times the size of our planned production library. Thus, for a 5000 member production library, a member virtual library would be enumerated as an SD file. The SD file is imported into the molecular modeling program Molecular Operating Environment (MOE), where the structures are rendered and physico-chemical properties were calculated and examined. Several iterations of this process usually were required to arrive upon the final set of reagents for a production library. An integral part of this process is running validation experiments with the chosen set of building blocks. Combining the results of validation studies with the building block choices derived from the desired set of physical-chemical properties yielded the final design of the production library. An advantage to this iterative approach is the high level of success in these production libraries: >90% of the compounds synthesized are produced with satisfactory yield and >80% pure. Continuing with the data management strategy, the final production set of structures were enumerated and associated with a unique ID number. This information, as well as the physico-chemical properties calculated for each compound, was loaded into our HTOS data system before synthesis takes place. Also at this time, the production run file for the IRORI system was generated and quantities of reagents needed for the production campaign are calculated using the SAR feature of ISIS. We maintained the production information management system separate and distinct from the HTOS database. The only information that needed to be transferred from the production management software back to the data system was the ID and archive location for that ID. All other aspects of the process were

3 Figure 3. Stepwise results handled by the IRORI AccuTag Synthesis Manager. Initially, MiniKans were loaded with the appropriate resin using a manual 96-channel dry resin loader. RF tags were also added to the MiniKans prior to capping. The split and mix process was managed by the IRORI AccuTag Synthesis Manager and the automated sorting was carried out on the IRORI AutoSort 10Kx instrument, which is designed to accept 20 Mettler-Toledo Bohdan MiniBlocks. Chemical steps were performed using standard synthesis glassware; we found no need for additional automation to support the chemistry steps. Washing steps were accomplished using either a manual resin washing apparatus or the automated IRORI wash station. This process was repeated until the series of chemical synthesis steps was completed. At this point, the MiniKans were sorted into 48-position Bohdan MiniBlocks in preparation for cleavage from the solid phase resin. Cleavage solvents can be added to the Mettler-Toledo Bohdan MiniBlocks using a 48-channel manifold attached to a Hamilton 500 solvent delivery system. The MiniBlocks were then subjected to incubation with the appropriate cleavage solvent while on a high capacity (6 block) shaking and washing station. Finally, the cleavage solution was collected by vacuum filtration into 12 X 75 glass test tubes arrayed in 48-position racks with a microtiter plate footprint. The cleavage solvent was removed by vacuum centrifugation and the dried samples are ready for reformatting and chemical analysis. The two-paired 48-position racks containing the dried down samples were solubilized using volatile organic solvents (mixtures of acetonitrile/methanol/water have given excellent results with Molecumetics libraries). The amount of solvent added should have yielded a 10mM solution of the compound based upon a calculated 80% overall yield for the chemistry performed. This expected yield was determined in the prior validation studies. The 10mM solution was transferred to a number of 96 well plates including several master source plates, a plate for LC/MS analysis and a plate for NMR analysis. The master source plates are dried down for long-term storage at 4 o C. The plate for LC/MS analysis consisted of 10ul of the 10mM compound solution diluted to 500ul with a 1:1 mixture of acetonitrile:water. This plate is then analyzed using a Micromass Platform LCZ API mass spectrometer outfitted with a Gilson 215 (Gilson, Middleton, WI) autosampler. Analytical liquid chromatography was performed on a Hewlett Packard 1100 HPLC system. Several detection methods in addition to the MS were employed including diode array UV detection and Evaporative Light Scattering Detection (ELSD).

4 Figure 4. Overall Results The column used was a Zorbax SB-C18 (2.1 X 30mm, 3.5u particle size) with a four minute run time consisting of a gradient of 5-95% acetonitrile in water (0.1% formic acid) over 2 minutes. Run time was 5.5 hours per plate. Data from the LC/MS run consisting of sample ID, identity, and purity information is parsed out of the Openlynx report file and imported into an Oracle table. Routines are run which automatically flag pass those compounds which exhibit the predicted mass and show a ELSD signal for a particular peak > 80%. The data for all compounds that fail this criteria are manually inspected: if the LC/MS run is unacceptable, these samples are resubmitted. If the LC/MS data clearly shows a synthesis failure, these samples are labeled fail in the database. Purification is not attempted on impure samples. The plate for NMR analysis consisted of 500 ul of the 10mM solution of compound in protic solvent. The plate was subjected to NMR analysis using a Varian VAST system consisting of a Gilson 215 liquid handler, a 120 ul NMR flow probe, and a Varian Mercury 400 MHz NMR spectrometer. Samples were injected from the plate into the flow probe directly using the liquid handler. NMR spectra were then acquired using scout-scan and WET solvent suppression; data collection takes place using 32 scans. Run time was 6-7 hours per plate. After data collection, the entire sample was recovered from the probe and collected

5 Miniblock containing Irori Kans and associated collection rack with microtiter plate footprint Bohdan 48 channel solvent delivery system back into its original source, the probe was washed, and the sequence was repeated. RESULTS Data analysis consists of performing automated routines that evaluate the results of the LC/MS data as well as the NMR information. After analytical data on an entire library were collected and evaluated, the success of the production campaign could be examined. Typically, a stepwise success rate was calculated, uncovering trends that could be corrected in future libraries. The overall synthesis success rate could also be determined. SUMMARY The successful production of high quality, individually characterized libraries of compounds has been demonstrated resulting from a well-developed workflow from library planning to synthesis through analytical characterization. Information management capabilities included the ability to generate virtual library, production libraries, and the ability to calculate the needed amounts of reagents and solvents for a production campaign. All compounds were registered into the library database allowing for tracking of production success and failure rates. Production of libraries utilized the IRORI split and mix system. Chemistry was performed using standard laboratory glassware. Single compounds per well were obtained after arraying the MiniKans into Mettler- Toledo Bohdan MiniBlocks and treatment with the appropriate cleavage cocktail. Collection into 48-position racks followed cleavage solution removal through vacuum centrifugation. Once the samples were reformatted into 96 well plates they were subsequently analyzed using high throughput flow injection NMR and LC/MS. Purity and yield were determined, and this information was recorded electronically into the database. Finally, acceptable compounds were reformatted from the original source plates into master plates and made available for testing. ACKNOWLEDGEMENTS The author wishes to thank the entire HTOS team at Molecumetics for their contributions to the success of this program. Individual contributors include: Chris Lum, Vince Huber, and Gangdahar Nagula for their chemistry support and validation,. Mark Deiparine and Chance Elliott for automation support, and Jim Dattilio, Stephanie Biegel-Orme, Dan Cunningham, Jennifer Martinez and Jennifer Young for production support. We also wish to thank Tomas Vaiser and Min Lee for LC/MS and NMR support. In addition, the IT staff, led by Mark Runyan, did an excellent job of supporting the information management strategy for the HTOS group. REFERENCES For a comprehensive review of chemistry performed using the IRORI system see: A.J. Mendonca, X.-y. Xiao IRORI technical publication Solid Phase Organic Synthesis using Microkan and MicroTube reactors Volume 1, 1998, Published by IRORI, Chemistry Developed at Molecumetics and applied to library production: M. Kahn et al Highly Efficient and Versatile Synthesis of Libraries of Constrained b-strand Mimetics, Biorg. and Med. Chem. Lett., (8) , 1998.