DRUG
`DEVELOPMENT
`
`ERIC D. CARLSON
`PEIJUN CONG
`WILLIAM H. CHANDLER Jr.
`HENRY K. CHAU
`THOMAS CREVIER
`PETER J. DESROSIERS *
`ROBERT D. DOOLEN
`CHRIS FREITAG
`LAUREN ATAGI HALL
`THOMAS KUDLA
`ROLLAND LUO
`COLIN MASUI
`JON ROGERS
`LI SONG
`ANNY TANGKILISAN
`KAREN QUYEN UNG
`LUPING WU
`
`Symyx Technologies Inc.
`3100 Central Expressway
`Santa Clara, CA 95051, USA
`
`* Corresponding author
`
`published by ~" srl
`Via Cesare da Sesto, 10
`20123 Milano (Italy)
`Tel. 0039 02 83241119
`0039 02 83241358
`Fax 0039 02 8376457
`
`An integrated
`high throughput
`workflow
`for pre-formulations:
`Polymorph and salt
`selection studies
`
`INTRODUCTION
`
`Pre-formulation, the selection of the
`form of an active pharmaceutical
`ingredient (API) most suitable for
`formulation, is a critical step in the drug
`development process. Recently, there
`has been considerable interest in
`reducing both the amount of time and
`the quantity of API needed to complete
`these studies including the use of high
`throughput techniques (1). One of the
`earliest reports of the use of high
`throughput crystallization of
`pharmaceuticals describes techniques
`for preparing arrays of salts in microtiter
`plates and screening for crystallinity
`using microscopy (2). More recently,
`reports describing the application of
`spectroscopic techniques to screen
`arrays of crystals in order to identify
`polymorphs have appeared (5). There
`are also reports of interesting techniques
`for preparing arrays of crystals (4)
`including polymer libraries (5), crystal
`nucleating chips (6), microfluidic devices
`(7), and acoustic injection of fluid
`droplets (8). In addition, methods
`describing the use of software for
`planning and analyzing the results of
`high throughput solid form screening
`have also appeared (9). Although each
`of these methods increases the speed
`or efficiency of individual steps in the
`pre-formulation process, none of these
`provide a truly integrated solution.
`Symyx has been developing high
`throughput techniques for material
`science since 1995 and has created
`integrated workflows in such diverse
`areas as homogeneous catalysis,
`heterogeneous catalysis, polymer
`formulations, pigments, electronic
`materials, and nano-dispersions (10). An
`important part of all these workflows is
`an integrated hardware and software
`system that allows the throughput at
`
`each of the key steps in a process, (e.g.,
`synthesis, analysis, data storage, data
`retrieval, data manipulation, and
`reporting, etc.), to be matched thereby
`eliminating bottlenecks. Working in
`collaboration with Merck, we have
`recently developed a high throughout
`workflow that allows crystallization, salt
`selection, and polymorph studies to be
`completed in a fraction of the time and
`using a smaller quantity of API than
`conventional studies (11).
`
`LIBRARY DESIGN
`
`An essential part of a successful high
`throughput workflow is the rational
`design of libraries that cover as wide a
`range of variables as possible, followed
`by additional experiments on those
`variables that are found to be critical to a
`particular API. For both salt selection and
`polymorph studies the goal is to prepare
`and characterization as many crystalline
`forms of an API as possible. Because
`finding crystallization conditions that
`result in the formation of different forms
`is essential to both types of studies, the
`initial experiments should cover a broad
`range of crystallization conditions,
`including method (e.g., cooling,
`evaporation, precipitation, and slurry),
`conditions (e.g., time, temperature,
`rate), solvent, and in the case of salt
`selection studies, counter-ion. Varying
`counter-ion in one Cartesian coordinate
`and crystallization solvent in the other
`dimension of an 8 x 12 array generates
`96 unique compositions that can be
`daughtered to allow three different sets
`of crystallizations conditions to be
`explored simultaneously (5 x 96
`crystallizations = 288 crystallizations/
`design). For polymorph studies
`compositional diversity is generated
`using solvent mixtures (Figure 1). Library
`
`JULY/AUGUST 2003
`
`DRUG DEVELOPMENT
`
`Lupin Ex. 1072 (Page 1 of 6)
`
`

`
`designs are constructed using
`Symyx’ Library Studio® a
`proprietary software program that
`allows the creation of complicated
`designs in minutes and stores to a
`database maps that describe the
`composition of each element and
`instructions that can be read by
`robotic automation software to
`execute the dispensing, heating,
`stirring, cooling specified in the
`design.
`For both polymorph and salt
`selection designs, the initial choice
`of solvents is critical. Although
`extensive literature exists
`describing methods for the
`classification of solvents for
`chromatography including the use
`of principal component analysis to
`assess properties such as solvent
`polarity (12), polarity is just one of
`the variables known to effect
`crystallization. To address this, a
`table consisting of over a hundred
`solvents and numerical values of
`fifteen of their physical properties
`was compiled. Standard statistical
`clustering techniques were then
`used to create a hierarchy of
`solvent relationships from which it
`was possible to divide the solvents
`into any number of groups. For a
`given crystallization design
`requiring n solvents, the solvents are
`clustered into n groups and one solvent
`is selected from each different group.
`
`Figure 1 - A Library Studio~ design used for polymorph studies contains twenty different solvents that are
`distributed to the master plate containing the API, equilibrated, filtered, and then daughtered to three different
`crystallizer assemblies allowing 96 evaporations, 96 precipitations, and 96 cooling crystallizations to be
`completed in a single day
`
`birefringence and Raman measurements
`are commercially available, none is
`compatible with XRD since the depth of
`the wells prevented reflectance XRD
`measurements. The solution was to
`create an array of vessels by sealing a
`flat glass substrate to a Teflon coated
`block as shown in Figure 2. The
`crystallizer assembly functions as 96
`isolated vessels that can be heated or
`cooled. After removal of the supernatant
`from the crystallizer assembly, the Teflon
`block can be removed leaving the
`crystals on a flat glass substrate.
`The filtration and crystallization
`assemblies fit into a series of heating
`and cooling devices on the deck of a
`liquid handling robot (Figure 2).
`Crystallization Station% proprietary
`Symyx software, allows the entire
`
`crystallization procedure including
`dispensing the crystallization solvents to
`an array of vials containing the API,
`heating the mixtures of API and
`crystallization solvents, filtering the hot
`solutions, daughtering the hot filtrates to
`three different crystallizer assemblies,
`cooling hot solutions, removing the
`solvents, sampling of aliquots for
`solubility measurements, and creation
`of a log file to be fully automated. A
`CCD camera with a telecentric lens
`mounted on a carriage underneath
`the crystallizer assemblies allows in situ
`birefringence images
`of the crystals formed
`to be obtained before the crystallizer
`assemblies are opened, facilitating the
`detection of unstable forms. Although
`the entire crystallization procedure
`requires 16 hours, the
`level of automation is
`sufficiently high it
`requires less than one
`hour of user time to
`carry out several
`hundred crystallizations.
`
`CRYSTALLIZATION
`
`Liquid handling robots suitable for
`dispensing solvents, heating,
`daughtering, and cooling arrays of
`solutions are readily available. There
`were, however, at least two significant
`challenges to developing a system for
`automated crystallizations. The first was
`hot filtration. All of the commercially
`available equipment for parallel filtration
`was deemed unsuitable for parallel
`crystallizations with multiple solvents
`due to a common headspace. A unique
`design developed for this workflow
`made it possible to create
`96 isolated filters from a
`single sheet of filtration
`media, while at the same
`time eliminating cross talk in
`the vapor phase. The
`second challenge was to
`form crystals on a substrate
`that would allow all of the
`analyses to be performed
`without manual
`manipulation of the
`crystalline samples.
`Although optically
`transparent crystallization
`vessels compatible with
`
`Figure 2 - The crystallizer assembly shown in expanded and assembled views
`provides a means of growing crystals that allows bffefringence, Raman, XRD,
`and melting point measurements to be made without manual manipulation
`
`SCREENING
`
`The selection of
`analytical techniques is
`critical to an integrated
`high throughput workflow.
`In addition to requiring
`
`DRUG DEVELOPMENT
`
`JULY/AUGUST 2003
`
`Lupin Ex. 1072 (Page 2 of 6)
`
`

`
`Figure 3 - Raman data for cimetidine sorted using Spectra StudioTM and
`displayed in library format shows three distinct groups. The starting form
`is shown in blue. Two other forms that were shown to be polymorphs are
`shown in green and yellow.
`
`the use of rapid serial or
`parallel analytical techniques to
`match the throughput of the
`crystallization station
`(288 crystallizations/day), it is
`essential that the techniques
`are capable of detecting and
`distinguishing different
`crystalline forms of the same
`composition. Although some
`pre-formulation systems might
`rely solely on Raman or XRD to
`screen for polymorphs, this
`was considered and deemed
`inadequate because no single
`analytical technique can be
`used to distinguish all unique
`crystalline forms. Thus, five
`complimentary analytical
`techniques were selected:
`solubility, birefringence,
`Raman, XRD, and melting
`point.
`Solubility measurements are
`obtained by drawing aliquots from the
`solutions at various points during the
`crystallization. Removal of the solvents,
`dilution, and subsequent LC analysis
`allows the concentration and stability of
`the API in all of the solvents used in the
`design to be determined in a few hours.
`In practice the solutions are sampled
`immediately after filtration, providing the
`concentration of the API or its salts at
`high temperature, and at the end of the
`cooling cycle providing the concentration
`at low temperature. EpochTM software
`controls the dilution, data acquisition,
`data analysis, and stores the data and
`acquisition parameters to the database.
`Bireffingence measurements are
`used to distinguish between crystalline
`and amorphous material. Higher
`resolution images are used to determine
`crystal habit and size. EpochTM software
`controls the acquisition and stores the
`images to the database. The
`bireffingence data for a set of 288
`crystallizations can be acquired in less
`than three hours and requires less than
`twenty minutes of user time.
`Raman spectra are acquired with a
`dispersive Raman microscope to
`distinguish different solid forms.
`Although several commercial
`instruments equipped with XY stages are
`suitable for analyzing libraries of crystals,
`software was needed to achieve the
`desired goal of an integrated workflow
`as the acquisition and analysis of
`hundred of Raman using the
`commercially available hard and
`software was prohibitively time
`consuming. Because the Raman focuses
`on single crystals rather than the entire
`sample, EpochTM software is used to
`facilitate the location of crystals and the
`creation of a map, Le., a list of
`coordinates identifying the location of
`the crystals of interest. Once the
`
`mapping and sorting.
`EpochTM software facilitates
`the location of crystals,
`create maps, controls the
`data acquisition, and stores
`the images, coordinates,
`area plots, and 2 theta
`plots to the database. The
`time used to acquire an
`XRD pattern varies with
`sample size, but the
`hardware used in this
`workflow allows adequate
`signal to noise ratios to be
`obtained on samples as
`small as 100 ~g in as little
`as three to five minutes,
`allowing the data from an
`entire set of 288
`crystallizations to be
`acquired in less than a day.
`Spectra StudioTM is
`used to sort the XRD data
`as described above for Raman. Due to
`the small sample size, non-statistical
`distributions of crystal orientations
`often result in dramatic changes in
`intensity for crystals of the same form
`rendering the direct XY correlation
`ineffective for sorting. The software
`was modified to allow the XRD
`patterns to be displayed as line
`spectra and then correlated based on
`peak positions.
`Even XRD data are insufficient to
`distinguish all unique crystalline forms of
`an API. Iso-structural solvates in
`particular can have indistinguishable
`Raman and XRD patterns. Moreover,
`although Raman and XRD can be used
`to distinguish different forms, they
`provide no information concerning the
`identity of these forms, e.g., solvates,
`hydrates, or true polymorphs, or their
`relative stability. DSC, TGA, and NMR are
`often used to provide this additional
`information in conventional studies, but
`technical challenges associated with
`making these measurements rapid serial
`or parallel caused us to explore other
`techniques. Using proprietary parallel
`birefringence technology, a parallel
`melting device was constructed. Initially
`designed to determine the melting point
`by recording the temperature at which
`the bireffingence signal dropped to zero
`as the crystals melted, the equipment
`developed can also detect changes in
`bireffingence attributable to solid-solid
`transitions such as polymorphic phase
`transitions and/or desolvation events.
`Typically run from 40 to 240°C at a
`ramp rate of 1 °C/minute, a run is
`complete in less than four hours
`allowing data from an entire set of
`crystallizations to be acquired in less
`than one day. EpochTM software controls
`the acquisition, data analysis, and stores
`traces of intensity versus temperature to
`the database.
`
`coordinates and the corresponding
`images for an entire library are stored to
`the database, the software acquires the
`Raman spectra for an entire library in a
`fully automated manner and stores the
`spectra and acquisition parameters to
`the database. Sorting the tens to
`hundreds of spectra generated each day
`is accomplished using Spectra StudioT%
`a proprietary Symyx software program,
`that allows the chemist to load hundreds
`or even thousands of Raman spectra
`from the database, enter a correlation
`factor, and then sort the spectra into
`groups such that members of the same
`group have a higher correlation than the
`user-defined value. To the extent that
`the decision of whether any two spectra
`are the same or different is ultimately a
`subjective one, the software has a
`number of features that allow the user
`to overlay spectra, merge and split
`groups, reassign spectra, and also to
`display the groups as they appear in
`library format (Figure 5). This software
`reduces the time required to sort the
`Raman spectra obtained from a set or
`sets of crystallizations from several days
`to a few hours.
`Although Raman is exceedingly
`sensitive (samples as small as 5 to
`10 Bg are sufficient) and the Raman
`spectra of the unique crystalline forms of
`an API are usually sufficiently different to
`be distinguished, this is not always the
`case. Indeed, we have encountered
`examples of unique crystalline forms
`that give indistinguishable Raman
`spectra. In addition, a small but non-
`negligible number of compounds
`fluoresce, thus requiring the use of other
`analytical techniques. XRD is inherently
`sensitive to changes in crystal packing.
`As was noted above for Raman, XRD
`instruments equipped with XY stages are
`commercially available although new
`software was needed to facilitate
`
`12 ~:;,.,m~i.’il~ JULY/AUGUST 2003
`
`DRUG DEVELOPMENT
`
`Lupin Ex. 1072 (Page 3 of 6)
`
`

`
`POLYMORPH STUDIES
`
`Cimetidine (1.00 g) was dissolved in
`methylene chloride (20 mL) and
`aliquots (200 ~L, 10 mg cimetidine)
`were dispensed to an 8 x 12 array of
`1 mL glass vials in an aluminum block.
`The solvent was removed by
`evaporation and the array was then
`transferred to the deck of the
`crystallization station and sealed with a
`gasket covered by a stainless steel cover
`with holes that allows liquid to be added
`and removed from the vials via a
`piercing needle. The crystallization
`procedure was executed using
`Crystallization StationTM that first
`prompted the user for the Library
`Studio® design number and the
`locations on the deck where
`the three crystallizer
`assemblies, two LC vial
`arrays, and 25
`recrystallization solvents
`specified in the design were
`placed. The program then
`controlled the dispense of
`the crystallization solvents
`(800 ~L/well) into the array
`of sealed vials containing
`cimetidine according the
`design shown in Figure 1
`using a single tipped
`piercing needle. The
`program then activated
`heaters and the array of
`cimetidine and
`recrystallization solvents
`along with the filtration
`assembly, the four-tipped
`needle, and the crystallizer
`assembly used for the
`cooling crystallization was
`brought to 65°C and
`allowed to equilibrate for
`two hours. During this time
`the program dispensed anti-
`solvents (500 ~L/well) to
`the sealed crystallizer
`assembly to be used for
`precipitations. After equilibration, the
`program executed a filter daughter
`sequence. Aliquots of the hot mixtures
`of cimetidine in the various
`recrystallization solvents (650
`were aspirated and then passed through
`the filters of the filtration assembly into
`an array of vials using the four-tipped
`piercing needle. The needles were
`washed to remove any traces of solids
`that may act as seeds, then used to
`withdraw the hot filtrates (550
`and to dispense aliquots to an open
`crystallizer for evaporative crystallization
`(200 ~L/well), to the sealed crystallizer
`containing anti-solvents for precipitations
`(1 O0 ~L/well), to the sealed crystallizer
`heated to 65°C for cooling
`crystallizations (200 ~L/well), and to the
`first array of LC sample vials
`
`transferred to the auto sampler of
`Agilent 11 O0 LC and EpocMM was used
`to execute a sequence that allowed for
`the collection, analysis, and storage to
`the database of LC data for each of the
`samples. The first data set that was used
`to determine the solubility at the
`beginning of the crystallization (65°C)
`and the second set to determine the
`solubility at the end of the crystallization
`0 o°c).
`The three universal substrates from
`the crystallizations were then analyzed
`using birefringence, Raman, XRD, and
`melting point. The birefringence images
`were obtained the same day the
`crystallizer assemblies were taken apart.
`EpochTM software prompted the user for
`the library number associated with each
`of the three substrates
`and then controlled the
`acquisition and stored the
`images to the database.
`The acquisition for each
`library was completely
`automated and required
`approximately
`45 minutes/library.
`While the
`birefringence images
`were being acquired on
`the second universal
`substrate, the first
`universal substrate was
`placed on the stage of a
`Jobin Yvon Raman
`spectrometer. EpochTM
`software controlled
`movement of the stage
`and after locating crystals
`suitable for Raman
`acquisition, the software
`stored coordinates and
`images of the crystals to
`the database. The
`software controlled the
`acquisition of data using
`the positions stored on
`the database and stored
`Raman spectra and
`acquisition parameters to the database.
`The process was then repeated for the
`remaining two universal substrates. In
`total, 147 Raman spectra were acquired
`over the course of two days and
`required less than three hours of user
`time. The Raman data were then sorted
`using Spectra StudioTM in less than two
`hours and resulted in the identification
`of three unique types of Raman, one of
`which was the same as the commercial
`material (Figure 5).
`Similar methods were used to obtain
`XRD data on a Bruker DX diffractometer
`and sorting of the XRD data using
`Spectra StudioTM confirmed the presence
`of three unique crystalline forms. Using
`the crystallization conditions from the
`automated crystallization experiments,
`samples of each of the three forms was
`
`(50 ~L/well). The entire filter daughter
`sequence was complete in less than two
`hours. The program then executed the
`controlled cooling cycle specified in the
`Library Studio® design, 65 to 10% over
`8 hours. After equilibrating for one hour
`at 10°C, the four-tipped needle
`aspirated aliquots of supematant from
`the cooling crystallizations and
`dispensed them to the second array of
`LC sample vials (50 ~L/well). Finally, the
`remaining supernatants were aspirated
`from the precipitation and cooling
`crystallizations and dispensed to a third
`array of vials for recovery.
`The following day, a camera
`mounted on a carriage underneath the
`crystallizer assemblies and controlled via
`software was used to acquire
`
`birefringence images of the 288
`crystallizations, two rows at a time. The
`covers and gaskets were then removed
`from the precipitation and cooling
`crystallizers and twelve prong wicks were
`inserted across rows to prevent the last
`traces of solvent from evaporating on
`the crystals. After drying in air for four
`hours, the crystallizer assemblies were
`taken apart and the three universal
`substrates with crystals were removed
`and stored in cases for subsequent
`analysis.
`The two arrays of LC vials were
`placed in a Genevac to remove the last
`traces of solvent, and aliquots of
`acetonitrile (500 ~L/well) were added
`as a dilution solvent using the liquid
`handling robot of the solubility station.
`After sealing and shaking, the vials were
`
`Figure 4 - The Perellel Melting Point StetionTM used to enelyze libreries of
`semples (upper/eft) meesures chenges in the intensity of e birefringence signel
`es e function of tempereture (lower/eft). Treces corresponding to the semples
`conteining pure Form I (upper right), pure Form II (middle right) end e mixture
`of the two forms (lower right) obteined during e polymorph study of
`nebumetone.
`
`DRUG DEVELOPMENT
`
`JULY/AUGUST 2003
`
`Lupin Ex. 1072 (Page 4 of 6)
`
`

`
`prepared on 50 mg
`scale and completely
`characterization by
`DSC, TGA, and NMR.
`Comparison of this
`data to the literature
`confirmed that the
`three groups identified
`from a single set of
`fully automated
`crystallizations and high
`throughput screening
`were true polymorphs
`of cimetidine (1
`When an identical
`set of crystallizations
`was carried out using
`nabumetone (1.00 g,
`3.3 mg/crystallization)
`Raman and XRD data
`were consistent with
`the presence of two
`forms. The melting
`points determined
`using the parallel
`melting point device
`agree to within a
`degree of the melting
`points reported
`previously for the two polymorphs of
`nabumetone (14). As shown in
`Figure 4, the presence of mixtures can
`be readily detected using this
`technology. The formation and
`characterization of Form II is particularly
`noteworthy because it is unstable with
`respect to conversion to Form I and was
`first observed from capillary tube
`crystallizations.
`
`SALT SELECTION STUDIES
`
`Figure 5. Raman and XRD screening
`completed within days of the
`crystallization resulted in the
`identification of eight crystalline salt
`forms of naproxen including both
`hydrated and anhydrous forms of the
`sodium salt and two polymorphs of the
`tromethamine salt. A similar set of
`experiments carried out using ephedrine
`(5 rag/well) and eleven acids distributed
`by column and eight crystallization
`solvents distributed by row resulted in
`the identification of seven crystalline
`salts of ephedrine.
`
`Figure 5 - The bbefringence images taken during a salt selection study on naproxen
`illustrates the quantity and quafity of crystals that may be obtained in a completely
`automated manner.
`
`That three
`polymorphs of
`cimetidine, the two
`polymorphs of
`nabumetone
`including an
`unstable form,
`eight salts of
`naproxen and
`seven salts of
`ephedrine were
`identified from four
`sets of
`crystallizations each
`using a gram or
`less of API
`illustrates the
`usefulness of the
`high throughput
`approach. The true
`value of the
`integrated system,
`however, only
`becomes apparent
`when it is realized
`that although each
`of the studies
`described above
`required a week,
`because each piece of hardware,
`including the crystallizer was used just
`one day during a given set of
`experiments, studies may be run
`concurrently allowing five such studies to
`be complete every week,
`
`PRE-FORMULATIONS
`DISCOVERY TOOL
`
`The hardware and software
`components along with proprietary
`searching (Polwiew~"), browsing
`(Renaissance Web
`Browseff"), and
`reporting software
`(Polymorph Reporting
`TooF"), comprise an
`integrated system for
`pre-formulations
`referred to as the
`Pre-formulations
`Discovery Tool
`(Figure 6). Working in
`collaboration with ten
`of the world’s leading
`pharmaceutical
`companies, including
`Merck and Eli Lilly who
`have since purchased
`this system, a team of
`six chemists has
`performed over 75,000
`crystallizations
`encompassing
`crystallization, salt
`selection, and
`polymorph studies on
`more than 70
`pharmaceutically active
`
`I~ispensing solutions of
`either acids or bases
`across rows or columns
`of an array of vials
`containing API followed
`by reaction and removal
`of solvents, generates
`arrays of crude salts.
`Recrystallization and
`analysis using the same
`procedure described
`above for the polymorph
`studies makes it possible
`to rapidly identify
`crystalline salt forms. For
`example, the reaction of
`naproxen (10.0 rag/well)
`with stoichiometric
`amounts of seven
`different bases dispensed
`across rows and
`crystallization from twelve
`different solvents
`dispensed down columns
`using the procedure
`described above yields
`the crystals shown in
`
`Figure 6 - The Pre-formulations Disocovery ToolTM System consists of seven stations
`linked by the Renaissance software suite and constitutes an integrated system
`capable of screening more than 50 API/year for salts and polymorphs.
`
`JULY/AUGUST 2003
`
`DRUG DEVELOPMENT
`
`Lupin Ex. 1072 (Page 5 of 6)
`
`

`
`compounds and resulting in the
`identification of over 500 unique
`crystalline forms in less than 18 months.
`
`ACKNOWLEDGEMENTS
`We greatly acknowledge the numerous
`contributions of our collaborators and
`are particularly grateful for the insights of
`Amin Khan, Stephen Maple, David
`Mathre, Paul Reider, Rick Sidler, Kara
`Sommervile and Greg Stephenson. We
`also acknowledge the engineering and
`software support provided by James
`Bennett, David Dorsett, AI Gushurt,
`Dennis Mullins, Tuyen Nguyen, and
`Lynn Van Erden as well as the useful
`feedback provided by the applications
`scientists Feresteh Lesani, Shah Lin,
`Helming Tan and Jonah Troth. Finally we
`wish to acknowledge Henry Weinberg
`for his useful conversations and support.
`
`REFERENCES
`
`1)
`
`NAKAGAMI, H. Farumashia 2003, 39 (3),
`
`204-208 and rel~rences therein
`
`2)
`
`BASTIN, N.J.; BOWKER, M J.; SLATER, B.J.
`
`Organic and Process Res. and
`
`3)
`
`4)
`
`Develop. 2000, 4, 427-435
`a. MORISSETTE, S.L.; SOUKASENE, S.;
`LEVINSON, D.; CIMA, M.J.; ALMARSSON,
`O. Proceediugs of the National
`Academy of Sciences of the l/uited
`States of America 2003, 100 (5),
`2180-2184 and references therein
`h. D’AMICO, K.L. PCT Int. AppL, 2001,
`39 pp. WO 01/82659 A1 20011101
`CARLTON, D.L.; DHINGRA, O.P.; IGO, D.H.;
`WATERS, P.W. (C[axo Croup Ltd., UK);
`PCT Int. AppL, 2000, 56 pp. WO 2000-
`US12610 20000509
`5) MATZGER, A.J.; LANG, M.; KIM, K. (The
`Regents of the University of Michigan,
`USA). PCT Int. AppL, 2003, 43 pp.
`WO 03/33462 A2 20030424
`LEE, T. U.S. Pat. AppL PuhL, 2003, 15 pp.
`US 2003068252 A1 20030410
`HANSEN, C.L.; QUAKE, S.R.; BERGEN, J.M.
`(California Institute of Technology, USA).
`U.S. Pat. AppL PuhL, 2003, 119 pp.;
`Cont.-in-part of U.S. Set. No. 887,997. US
`2003061687 A1 20030403 and rel~rences
`therein
`8) MUTZ, M.W.; ELLSON, R.N.; STEARNS, R.C.
`(Pico[iter In(:., USA). PCT Int. AppL,
`2002, 91 pp. WO 02/66713 A1 20020829
`LEVINSON, D.A.; CHIN, D. (USA). U.S. Pat.
`AppL PuhL, 2002, 29 pp. US 2002177167
`A1 20021128
`10) MURPHY, V.; TURNER, H.W.; WESKAMP, T.
`"High-throughput approaches to
`
`6)
`
`7)
`
`9)
`
`homogeneous catalysis" in Applied
`Homogeneous Catalysis with
`Organometa[[ic Compounds (2nd Edition)
`and rd’erences therein
`11) CARLSON, E.D.; CONG, P.; CHANDLER,
`W.H., JR.; CHAU, H.K.; DANIELSON, E.;
`DESROSIERS, P.J.; DOOLEN, R.D.; WU, L.
`(Symyx Technologies, Inc., USA). PCT Int.
`App[., 2003, 229 pp. WO 0314732
`A1 20030220
`12) a. SARBU, C. Revista de Chimie
`(Bucharest) 1999, 50 (12), 852-863
`h. J. of Chemical Iuf!~rmation and
`Computer Sciences 1999, 39 (4),
`692-698 and rel~rences therein
`c. BOTA, A.; SARBU, C.; MARUTOIU, C.;
`COMAN, V. Journal of Planar
`Chromatography - Modern TLC
`1997, 10 (5), 358-361 and rel~rences
`therein
`13) a. MIDDLETON, D.A.; LE DUFF, C.S.;
`PENG, X.; REID, D.C.; SAUNDERS,
`D. J.Am. Chem. Soc. 2000,
`122, 1161-1170 and references
`therein
`h. SHIBATA, M.; KOKUBO, H.; MORIMOTO,
`K.; MORISUAKI, K.; ISHIDA, T.; INIOUE,
`M. J. Pharma. Sciences 1983, 72
`(12), 1436-1442
`14) CHYALL, L.J.; TOWER, J.M.; COATES, D.A.;
`HOUSTON, T.L.; CHILDS, S.L. Crystal
`Growth and Design 2002, 2 (6),
`505-510
`
`DRUG DEVELOPMENT
`
`Lupin Ex. 1072 (Page 6 of 6)