Advertisement

Integrated and automated high-throughput purification of libraries on microscale

  • Carol Ginsburg-Moraff
    Correspondence
    Corresponding authors.
    Affiliations
    Global Discovery Chemistry, Novartis Institutes for Biomedical Research, Novartis Pharma AG., Cambridge, MA 02139, USA
    Search for articles by this author
  • Author Footnotes
    † Jonathan Grob: Valo Health, 75 Hayden Ave., Lexington, MA 02421, USA
    Jonathan Grob
    Footnotes
    † Jonathan Grob: Valo Health, 75 Hayden Ave., Lexington, MA 02421, USA
    Affiliations
    Global Discovery Chemistry, Novartis Institutes for Biomedical Research, Novartis Pharma AG., Cambridge, MA 02139, USA
    Search for articles by this author
  • Karl Chin
    Affiliations
    Global Discovery Chemistry, Novartis Institutes for Biomedical Research, Novartis Pharma AG., Cambridge, MA 02139, USA
    Search for articles by this author
  • Grant Eastman
    Affiliations
    Global Discovery Chemistry, Novartis Institutes for Biomedical Research, Novartis Pharma AG., Cambridge, MA 02139, USA
    Search for articles by this author
  • Sandra Wildhaber
    Affiliations
    Global Discovery Chemistry, Novartis Institutes for Biomedical Research, Novartis Pharma AG., Fabrikstrasse 1, Basel 4056, Switzerland
    Search for articles by this author
  • Mark Bayliss
    Affiliations
    Virscidian Inc., Cary, NC 27511, USA
    Search for articles by this author
  • Heinrich M. Mues
    Affiliations
    Global Discovery Chemistry, Novartis Institutes for Biomedical Research, Novartis Pharma AG., Fabrikstrasse 1, Basel 4056, Switzerland
    Search for articles by this author
  • Marco Palmieri
    Affiliations
    Global Discovery Chemistry, Novartis Institutes for Biomedical Research, Novartis Pharma AG., Fabrikstrasse 1, Basel 4056, Switzerland
    Search for articles by this author
  • Jennifer Poirier
    Affiliations
    Global Discovery Chemistry, Novartis Institutes for Biomedical Research, Novartis Pharma AG., Cambridge, MA 02139, USA
    Search for articles by this author
  • Marcel Reck
    Affiliations
    Global Discovery Chemistry, Novartis Institutes for Biomedical Research, Novartis Pharma AG., Fabrikstrasse 1, Basel 4056, Switzerland
    Search for articles by this author
  • Alexandre Luneau
    Affiliations
    Global Discovery Chemistry, Novartis Institutes for Biomedical Research, Novartis Pharma AG., Fabrikstrasse 1, Basel 4056, Switzerland
    Search for articles by this author
  • Stephane Rodde
    Affiliations
    Global Discovery Chemistry, Novartis Institutes for Biomedical Research, Novartis Pharma AG., Fabrikstrasse 1, Basel 4056, Switzerland
    Search for articles by this author
  • John Reilly
    Affiliations
    Global Discovery Chemistry, Novartis Institutes for Biomedical Research, Novartis Pharma AG., Fabrikstrasse 1, Basel 4056, Switzerland
    Search for articles by this author
  • Trixie Wagner
    Affiliations
    Global Discovery Chemistry, Novartis Institutes for Biomedical Research, Novartis Pharma AG., Fabrikstrasse 1, Basel 4056, Switzerland
    Search for articles by this author
  • Cara E. Brocklehurst
    Affiliations
    Global Discovery Chemistry, Novartis Institutes for Biomedical Research, Novartis Pharma AG., Fabrikstrasse 1, Basel 4056, Switzerland
    Search for articles by this author
  • René Wyler
    Affiliations
    Global Discovery Chemistry, Novartis Institutes for Biomedical Research, Novartis Pharma AG., Fabrikstrasse 1, Basel 4056, Switzerland
    Search for articles by this author
  • Author Footnotes
    § David Dunstan: Relay Therapeutics, 399 Binney Street, Cambridge, MA 02139, USA. Email: ddunstan@relaytx.com
    David Dunstan
    Correspondence
    Corresponding authors.
    Footnotes
    § David Dunstan: Relay Therapeutics, 399 Binney Street, Cambridge, MA 02139, USA. Email: ddunstan@relaytx.com
    Affiliations
    Global Discovery Chemistry, Novartis Institutes for Biomedical Research, Novartis Pharma AG., Cambridge, MA 02139, USA
    Search for articles by this author
  • Author Footnotes
    ‡ Alexander Marziale: AstraZeneca, Pepparedsleden 1, SE-431 83 Mölndal, Sweden. Email: alexander.marziale@astrazeneca.com
    Alexander N. Marziale
    Correspondence
    Corresponding authors.
    Footnotes
    ‡ Alexander Marziale: AstraZeneca, Pepparedsleden 1, SE-431 83 Mölndal, Sweden. Email: alexander.marziale@astrazeneca.com
    Affiliations
    Global Discovery Chemistry, Novartis Institutes for Biomedical Research, Novartis Pharma AG., Fabrikstrasse 1, Basel 4056, Switzerland
    Search for articles by this author
  • Author Footnotes
    † Jonathan Grob: Valo Health, 75 Hayden Ave., Lexington, MA 02421, USA
    § David Dunstan: Relay Therapeutics, 399 Binney Street, Cambridge, MA 02139, USA. Email: ddunstan@relaytx.com
    ‡ Alexander Marziale: AstraZeneca, Pepparedsleden 1, SE-431 83 Mölndal, Sweden. Email: alexander.marziale@astrazeneca.com
Open AccessPublished:August 23, 2022DOI:https://doi.org/10.1016/j.slast.2022.08.002

      Abstract

      We herein report the development of an automation platform for rapid purification and quantification of chemical libraries including reformatting of chemical matter to 10 mM DMSO stock solutions. This fully integrated workflow features tailored conditions for preparative reversed-phase (RP) HPLC-MS on microscale based on analytical data, online fraction QC and CAD-based quantification as well as automated reformatting to enable rapid purification of chemical libraries. This automated workflow is entirely solution-based, eliminating the need to weigh or handle solids. This increases process efficiency and creates a link between high-throughput synthesis and profiling of novel chemical matter with respect to biological and physicochemical properties in relevant assays.

      Keywords

      1. Introduction

      After a decade of stagnation, the field of chemistry automation for drug discovery is currently experiencing a renaissance, second only to the rise of combinatorial and parallel chemistry in the 1990s and 2000s.[
      • Schneider G.
      Automating drug discovery.
      ] This development is fueled by fundamental progress in automated synthesis,[
      • Peplow M.
      The robo-chemist.
      ,
      • Guetzoyan L
      • Ingham R.J.
      • Nikbin N.
      • Rossignol J.
      • Wolling M.
      • Baumert M.
      • Burgess-Brown N.A.
      • StrainDamerell C.M.
      • Shrestha L.
      • Brennan P.E.
      • Fedorov O.
      • Knapp S.
      • Ley S.V.
      Machine-assisted synthesis of modulators of the histone reader BRD9 using flow methods of chemistry and frontal affinity chromatography.
      ] reactor technologies,[
      • McMullen J.P.
      • Jensen K.F.
      Integrated microreactors for reaction automation: new approaches to reaction development.
      ,
      • Elvira K.S.
      • Casedevall i Solvas X.
      • Wootton R.C.R.
      • deMello A.J.
      The past, present and potential for microfluidic reactor technology in chemical synthesis.
      ,
      • Hwang Y.-J.
      • Coley C.W.
      • Abolhasani M.
      • Marzinzik A.L.
      • Koch G.
      • Spanka C.
      • Lehmann H.
      • Jensen K.F.
      Segmented flow platform for on-demand medicinal chemistry and compound synthesis: efficient reaction screening in oscillating droplets.
      ] integrated drug discovery,[
      • Michael Werner M.
      • Kuratli C.
      • Martin R.E.
      • Hochstrasser R.
      • Wechsler D.
      • Enderle T.
      • Alanine A.I.
      • Vogel H
      Seamless integration of dose-response screening and flow chemistry: efficient generation of structure–activity relationship data of β-Secretase (BACE1) inhibitors.
      ,
      • Czechtizky W.
      • Dedio J.
      • Desai B.
      • Dixon K.
      • Farrant E.
      • Feng Q.
      • Morgan T.
      • Parry D.M.
      • Ramjee M.K.
      • Selway C.N.
      • Schmidt T.
      • Tarver G.J.
      • Wright A.G.
      Integrated synthesis and testing of substituted xanthine based DPP4 inhibitors: application to drug discovery.
      ,
      • Chow S.
      • Liver S.
      • Nelson A.
      Streamlining bioactive molecular discovery through integration and automation.
      ,
      • Baranczak A.
      • Tu N.P.
      • Marjanovic J.
      • Searle P.A.
      • Vasudevan A.
      • Djuric S.W.
      Integrated platform for expedited synthesis-purification-testing of small molecule libraries.
      ] microscale chemistry[
      • Buitrago Santanilla A.
      • Regalado E.L.
      • Pereira T.
      • Shevlin M.
      • Bateman K.
      • Campeau L.-C.
      • Schneeweis J.
      • Berritt S.
      • Shi Z.-C.
      • Nantermet P.
      • Liu Y.
      • Helmy R.
      • Welch C.J.
      • Vachal P.
      • Davies I.W.
      • Cernak T.
      • Dreher S.D.
      Nanomole-scale high-throughput chemistry for the synthesis of complex molecules.
      ,
      • Mitchell M.C.
      • Spikmans V.
      • de Mello A.J.
      Microchip-based synthesis and analysis: control of multicomponent reaction products and intermediates.
      ,
      • Garcia-Egido E.
      • Spikmans V.
      • Wong S.Y.F.
      • Warrington B.H.
      Synthesis and analysis of combinatorial libraries performed in an automated micro reactor system.
      ,
      • Shaabani S.
      • Xu R.
      • Ahmadianmoghaddam M
      • Gao L.
      • Stahorsky M.
      • Olechno J.
      • Ellson R.
      • Kossenjans M.
      • Helan V.
      • Dömling A.
      Automated and accelerated synthesis of indole derivatives on a nanoscale.
      ] and high-throughput experimentation.[
      • Gesmundo N.J.
      • Sauvagnat B.
      • Curran P.J.
      • Richards M.P.
      • Andrews C.L.
      • Dandliker P.J.
      • Cernak T.
      Nanoscale synthesis and affinity ranking.
      ,
      • Osipyan A.
      • Shaabani S.
      • Warmerdam R.
      • Shishkina S.V.
      • Boltz H.
      • Dömling A.
      Automated, accelerated nanoscale synthesis of iminopyrrolidines.
      ,
      • Weller H.N.
      • Nirschl D.S.
      • Paulson J.L.
      • Hoffman S.L.
      • Bullock W.H.
      Addressing the medicinal chemistry bottleneck: a lean approach to centralized purification.
      ,
      • Liu M.
      • Kuanchang C.
      • Christian D.
      • Fatima T.
      • Pisarnitski N.
      • Streckfuss E.
      • Zhang C.
      • Xia L.
      • Borges S.
      • Shi Z.
      • Vachal P.
      • Tata J.
      • Athanasopoulos J.
      High-throughput purification platform in support of drug discovery.
      ] Miniaturization of synthetic chemistry has already enabled efficient optimization of reaction conditions for transformations relevant to medicinal chemistry while simultaneously decreasing the ecological footprint of such efforts.[
      • Wang Y.
      • Shaabani S.
      • Ahmadianmoghaddam M.
      • Gao L.
      • Xu R.
      • Kurpiewska K.
      • Kalinowska-Tluscik J.
      • Olechno J.
      • Ellson R.
      • Kossenjans M.
      • Helan V.
      • Groves M.
      • Dömling A.
      Acoustic droplet ejection enabled automated reaction scouting.
      ]
      What differentiates the current development from previous attempts is the full integration of key scientific disciplines in early hit-to-lead drug discovery. This includes synthesis, analytical chemistry, separation sciences, and physicochemical and biological profiling as well as characterization of ADME (absorption, distribution, metabolism, excretion) properties.[
      • Schneider G.
      Automating drug discovery.
      ,
      • Michael Werner M.
      • Kuratli C.
      • Martin R.E.
      • Hochstrasser R.
      • Wechsler D.
      • Enderle T.
      • Alanine A.I.
      • Vogel H
      Seamless integration of dose-response screening and flow chemistry: efficient generation of structure–activity relationship data of β-Secretase (BACE1) inhibitors.
      ,
      • Baranczak A.
      • Tu N.P.
      • Marjanovic J.
      • Searle P.A.
      • Vasudevan A.
      • Djuric S.W.
      Integrated platform for expedited synthesis-purification-testing of small molecule libraries.
      ,
      • Desai B.
      • Dixon K.
      • Farrant E.
      • Feng Q.
      • Gibson K.R.
      • van Hoorn W.P.
      • Mills J.
      • Morgan T.
      • Parry D.M.
      • Ramjee M.K.
      • Selway C.N.
      • Tarver G.J.
      • Whitlock G.
      • Wright A.G.
      Rapid discovery of a novel series of AbI kinase inhibitors by application of an integrated microfluidic synthesis and screening platform.
      ] This integration shifts the focus from generating chemical matter to establishing knowledge for early drug discovery projects. The plethora of data generated through integrated, platform-based approaches can only be managed and effectively utilized through data science techniques such as machine learning (ML).[
      • Gawehn E.
      • Hiss J.A.
      • Schneider G.
      Deep learning in drug discovery.
      ] These predictive ML models drive the multi-parameter optimization of chemical and property space in an iterative approach and their performance is fundamentally dependent on the quality of the data used to train them. In the context of integrated drug discovery, high quality data must be generated by testing well validated chemical matter in an array of robust assays.
      Novel techniques for hit-finding such as affinity selection mass spectroscopy (ASMS), differential scanning fluorimetry (DSF) or DNA encoded libraries (DEL) complement traditional high-throughput screening (HTS) approaches for the identification of chemical starting points for drug discovery. These hit-discovery approaches have matured to a state where the rate of false positive and false negative findings can be sufficiently mitigated.[
      • Gao K.
      • Shaabani S.
      • Xu R.
      • Zarganes-Tzitzikas T.
      • Gao L.
      • Ahmadianmoghaddam M.
      • Groves M.R.
      • Dömling A.
      Nanoscale, automated, high throughput synthesis and screening for the accelerated discovery of protein modifiers.
      ,
      • O'Connell T.N.
      • Ramsay J.
      • Rieth J.R.
      • Shapiro M.J.
      • Stroh J.G.
      Solution-based indirect affinity selection mass spectroscopy—a general tool for high-throughput screening of pharmaceutical compound libraries.
      ,
      • Gabriel J.
      • Höfner G.
      • Wanner K.T.
      A library screening strategy combining the concepts of MS binding assays and affinity selection mass spectrometry.
      ,
      • Qin S.
      • Meng M.
      • Yang D.
      • Bai W.
      • Lu Y.
      • Peng Y.
      • Song G.
      • Wu Y.
      • Zhou Q.
      • Zhao S.
      • Huang X.
      • McCorvy J.D.
      • Cai X.
      • Dai A.
      • Roth B.L.
      • Hanson M.A.
      • Liu Z.-J.
      • Wang M.-W.
      • Stevens R.C.
      • Shui W.
      High-throughput identification of G protein-coupled receptor modulators through affinity mass spectrometry screening.
      ,
      • Blakskjaer P.
      • Heitner T.
      • Hansen N.J.V.
      Fidelity by design: yoctoreactor and binder trap enrichment for small-molecule DNA-encoded libraries and drug discovery.
      ,
      • Satz A.L.
      • Kuai L.
      • Peng X.
      Selections and screenings of DNA-encoded chemical libraries against enzymes and cellular targets.
      ,
      • Reddavide F.V.
      • Cui M.
      • Lin W.
      • Fu N.
      • Heiden S.
      • Andrade H.
      • Thompson M.
      • Zhang Y.
      Second generation DNA-encoded dynamic combinatorial chemical libraries.
      ] However, more complex assay types (i.e., biochemical and cellular functional assays) require high quality chemicals. Residual reagents such as catalysts, bases, and byproducts could compromise assay read-outs. Approaches such as ASMS or DEL currently only yield qualitative biological data and fail to deliver reliable negative structure activity relationships (SAR). Data quality is paramount when generating quantitative SAR for hit-to-lead drug discovery to effectively drive optimization of a chemical series based on biological activity and physicochemical properties. This is reflected by the fact that the majority of automated platforms for drug discovery in industry and academia integrate a purification step.[
      • Michael Werner M.
      • Kuratli C.
      • Martin R.E.
      • Hochstrasser R.
      • Wechsler D.
      • Enderle T.
      • Alanine A.I.
      • Vogel H
      Seamless integration of dose-response screening and flow chemistry: efficient generation of structure–activity relationship data of β-Secretase (BACE1) inhibitors.
      ,
      • Dykstra K.D.
      • Streckfuss E.
      • Liu M.
      • Liu J.
      • Yu Y.
      • Wang M.
      • Kozlowski J.A.
      • Myers R.W.
      • Buevich A.V.
      • Maletic M.M.
      • Vachal P.
      • Krska S.W.
      Synthesis of HDAC inhibitor libraries via mircoscale workflow.
      ,
      • Hochlowski J.E.
      • Searle P.A.
      • Tu N.P.
      • Pan J.Y.
      • Spanton S.G.
      • Djuric S.W.
      An integrated synthesis-purification system to accelerate the generation of compounds in pharmaceutical discovery.
      ,
      • Sutherland J.D.
      • Tu N.P.
      • Nemcek T.A.
      • Searle P.A.
      • Hochlowski J.E.
      • Djuric S.W.
      • Pan J.Y.
      An automated synthesis-purification-sample-management platform for the accelerated generation of pharmaceutical candidates.
      ] Despite the revitalization of automation in medicinal and synthetic chemistry, the number of reports on fully automated modules for high-throughput purification and quantification integrating drying and reformatting steps remain scarce.
      We herein report the development and validation of an integrated and automated platform for high-throughput purification of chemical libraries on microscale (2-6 µmol per reaction). This platform integrates consecutive process steps from analytical UHPLC-MS (Ultra high-pressure liquid chromatography mass spectroscopy) based characterization of crude reaction mixtures to preparative purification, fractionation, quality control, quantification, drying and re-stocking to 10 mM DMSO stock solutions (Fig. 1). The automation of both sample and data handling enables rapid access to high-quality chemical matter and is a key factor in the downstream generation of profiling data which is critical to accelerating design-make-test-analyze (DMTA) cycles.
      Fig 1
      Fig. 1Integrated and automated purification of microscale libraries workflow.

      2. Materials & methods

      High-throughput synthesis and parallel chemistry are powerful tools in medicinal chemistry and have found numerous applications such as the rapid expansion of HTS hits, optimization of research compounds towards lead-like molecules and exploration of structure-activity relationships in chemical space.[
      • Lall M.S.
      • Bassyouni A.
      • Bradow J.
      • Brown M.
      • Bundesmann M.
      • Chen J.
      • Ciszewski G.
      • Hagen A.E.
      • Hyek D.
      • Jenkinson S.
      • Liu B.
      • Obach R.S.
      • Pan S.
      • Reilly U.
      • Sach N.
      • Smaltz D.J.
      • Spracklin D.K.
      • Starr J.
      • Wagenaar M.
      • Walker G.S.
      Late-stage lead diversification coupled with quantitative nuclear magnetic resonance spectroscopy to identify new structure–activity relationship vectors at nanomole-scale synthesis: application to loratadine, a human histamine H1 receptor inverse agonist.
      ] Subjecting large numbers of compounds produced by high-throughput synthesis workflows to downstream processes such as column chromatography, fractionation, drying, fraction pooling and quantification presents a significant technological and logistic challenge.[
      • Dykstra K.D.
      • Streckfuss E.
      • Liu M.
      • Liu J.
      • Yu Y.
      • Wang M.
      • Kozlowski J.A.
      • Myers R.W.
      • Buevich A.V.
      • Maletic M.M.
      • Vachal P.
      • Krska S.W.
      Synthesis of HDAC inhibitor libraries via mircoscale workflow.
      ] Traditional library workflows typically require fractionation into glass tubes due to the scale of synthesis and rely on gravimetric quantification of the isolated material. These processes typically involve two dry-down steps after both fractionation and pooling. Downscaling synthetic chemistry to a micromolar regime enables us to overcome these technical hurdles and represents a paradigm shift in library synthesis.[
      • Chow S.
      • Liver S.
      • Nelson A.
      Streamlining bioactive molecular discovery through integration and automation.
      ,
      • Barhate C.L.
      • Donnell A.F.
      • Davies M.
      • Li L.
      • Zhang Y.
      • Yang F.
      • Black R.
      • Zipp G.
      • Zhang Y.
      • Cavallaro C.
      • Priestly E.S.
      • Weller H.N.
      Microscale purification in support of high-throughput medicinal chemistry.
      ] The majority of assays require mere nanograms of compound for testing, meaning that traditional parallel and library chemistry workflows produce vast excesses of material for early hit-to-lead optimization.[
      • Ellson R.
      Picoliter: enabling precise transfer of nanoliter and picoliter volumes.
      ,
      • Dawes T.D.
      • Turincio R.
      • Jones S.W.
      • Rodriguez R.A.
      • Gadiagellan D.
      • Thana P.
      • Clark K.R.
      • Gustafson A.E.
      • Orren L.
      • Liimatta M.
      Compound transfer by acoustic droplet ejection promotes quality and efficiency in ultra-high-throughput screening campaigns.
      ] By generating material on a scale closer to that of biological assay profiling workflows, we can reduce the environmental impact of our chemical synthesis, and leverage biological automation systems using the ANSI/SLAS microplate standards plate format.[
      • Dykstra K.D.
      • Streckfuss E.
      • Liu M.
      • Liu J.
      • Yu Y.
      • Wang M.
      • Kozlowski J.A.
      • Myers R.W.
      • Buevich A.V.
      • Maletic M.M.
      • Vachal P.
      • Krska S.W.
      Synthesis of HDAC inhibitor libraries via mircoscale workflow.
      ,
      • Wang J.
      • Holloway T.
      • Lisova K.
      • van Dam R.M.
      Green and efficient synthesis of the radiopharmaceutical [18F]FDOPA using a microdroplet reactor.
      ]
      The maturation of library synthesis workflows on microscale at Novartis led us to pursue a de novo design for a suitable analysis and purification process to match the synthetic output. Library synthesis at Novartis’ Discovery Chemistry group is conducted in a 96-well microplate format on 2-6 µmol scale per well. The preparative chromatography step within this automation workflow is optimized to this synthetic scale with respect to the inner-diameter of the chromatography column, flow rate, tubing, splitter settings, injection loop volume and MS-sensitivity. For isolation of compounds from smaller reaction scales an analytical LC instrument is more suitable. While others within the pharmaceutical industry have demonstrated a further reduction in scale to the nanomolar regime,[
      • Buitrago Santanilla A.
      • Regalado E.L.
      • Pereira T.
      • Shevlin M.
      • Bateman K.
      • Campeau L.-C.
      • Schneeweis J.
      • Berritt S.
      • Shi Z.-C.
      • Nantermet P.
      • Liu Y.
      • Helmy R.
      • Welch C.J.
      • Vachal P.
      • Davies I.W.
      • Cernak T.
      • Dreher S.D.
      Nanomole-scale high-throughput chemistry for the synthesis of complex molecules.
      ,
      • Gesmundo N.J.
      • Sauvagnat B.
      • Curran P.J.
      • Richards M.P.
      • Andrews C.L.
      • Dandliker P.J.
      • Cernak T.
      Nanoscale synthesis and affinity ranking.
      ] this approach currently does not allow for isolation of the compound of interest and restocking to 10 mM DMSO stock for plating. Downstream profiling of nanoscale libraries is consequently limited to biophysical methods. The microscale workflow described herein aims to generate sufficient material to enable a serial dilution and delivery of assay-ready plates. The latter are applied in the context of a target-specific biochemical and cellular evaluation in addition to physicochemical and ADME profiling to generate a holistic data package for multi-parameter optimization. Furthermore, conducting synthesis on microscale enables the submission of residual DMSO stock to the Novartis compound library for future testing.
      Upon completion of synthesis and workup steps, a liquid handling robot transfers aliquots of the crude reaction mixtures into microtiter plates, and the plates are subjected to UHPLC-MS analysis. The Waters Acquity classic system (Waters Corp., Milford, MA, USA) is equipped with a diode-array detector (DAD), a QDa mass-spectrometer (MS) and a Corona Veo charged aerosol detector (CAD) (ThermoFisher Scientific, Waltham, MA, USA). All crude reactions undergo analysis under both high and low pH conditions on a BEH (ethylene bridged hybrid) C18 column using ammonium hydroxide and formic acid as modifiers respectively. All LC instruments are controlled by the Waters MassLynx software.
      We use Virscidian's Analytical Studio software (Virscidian Inc, Cary, NC, USA) to support and automate the analysis of liquid chromatography (LC) data. The samples are annotated with metadata prior to acquisition which enables compound specific data such as molecular formulas and molecular weights for the starting material, product of interest, byproducts, and internal standard to be carried through the workflow. This information is included in the MassLynx sample list prior to data acquisition. The annotation with metadata enables automated identification and highlighting of peaks in Analytical Studio during the automated purification process. To further facilitate the visualization of synthesis success across a plate, while simultaneously highlighting which crude mixtures are progressing to preparative HPLC, we have deployed a color-coded heat map as part of the user interface within Analytical Studio (Fig. 2). Reactions that demonstrate sufficient conversion to product for subsequent isolation and profiling are colored in either green or purple (the latter indicating potential co-elution of an impurity with the product of interest). Correspondingly, reactions with little or no conversion to product are colored in yellow or red respectively. This visualization allows the user to quickly identify reactions that require manual review, while minimizing time spent analyzing unambiguous results. The number of reactions that will require review on any given plate is often dependent on the type of chemical reaction and its substrate scope.
      Fig 2
      Fig. 2Analytical studio user interface featuring peak annotation and heat map visualization.
      The application of orthogonal liquid chromatography conditions (high and low pH modifiers in the mobile phase) in combination with the previously mentioned detectors provides a detailed understanding of the crude reaction profiles. The Analytical Studio software simultaneously resolves which reactions to purify by assessing the conversion to product and then identifies optimal LC conditions for subsequent preparative chromatography. Analytical Studio is equipped with decision-making capabilities that exclusively suggest reactions for preparative purification which display sufficient conversion to the desired product (>5% DAD or CAD peak area). The software furthermore derives and dynamically selects LC conditions for preparative HPLC from the analytical data by suggesting either generic or focused gradient programs. Chromatographic resolution for complex crude reaction mixtures is increased by applying a focused gradient that extends the duration during the target product elution window. A linear correlation between analytical retention times and focused gradient windows for preparative HPLC has been established by geometric scaling of the LC columns and conditions.[

      Aubin, A. J.; Cleary, R. Analytical HPLC to preparative HPLC: scale-up techniques using a natural product extract. Waters Application Note 720003120 2019.

      ,

      A table listing retention times and the corresponding focused gradient windows can be found in the supporting information.

      ] To assess the complexity of the crude reaction mixtures, Analytical Studio analyzes the MS trace by extracting a composite of known peaks of interest and comparing this to a composite of the unknown ions. The retention time and the resulting peak shape determine the overlap between the desired product mass with residual reagents or unknown byproducts. This allows the software to accurately determine closely eluting components that might not be apparent from TIC (Total Ion Chromatogram) or TAC (Total Absorbance Chromatogram) signals on the MS or DAD detector. In addition, Analytical Studio compares high and low pH chromatograms for each reaction mixture by analyzing LC parameters including resolution, retention time, peak shape, peak asymmetry, tailing etc., to determine a total method score for both conditions. The pH with the superior overall score is applied during preparative HPLC.
      Data acquisition and analysis of chromatograms, for a full library of 96 reactions is completed in about 3.5 h using one-minute gradients (2-98% acetonitrile in water). The Analytical Studio software processes this data and generates a worklist summarizing its decisions and informing preparative chromatography. While the process is fully automated, the operator can override decisions made by the software and add additional reactions to the sample list or adjust preparative chromatography conditions.
      Preparative HPLC is subsequently executed on a Waters AutoPurification system equipped with a QDa mass detector and a photodiode array detector, which is integrated into the “crude to DMSO stock” automation platform. This platform combines semi-preparative purification and fraction QC with quantification via a charged aerosol detector. Samples are dried through a Porvair Ultravap Mistral blow-down dryer (Porvair PLC, Norfolk, UK), and re-dissolution to 10 mM DMSO stock solutions using a Formulatrix Mantis liquid dispenser (Formulatrix, Bedford, MA, USA). A ThermoFisher Scientific Spinnaker (ThermoFisher Scientific, Waltham, MA, USA) robot, which is controlled by the ThermoFisher Momentum software, integrates the individual components of the automation platform (Fig. 3). Reversed-phase HPLC is performed at a flow rate of 10 mL/min on a 10 x 150 mm C18 XBridge column with 5 µm BEH particles.
      Fig 3
      Fig. 3A. Schematic representation of the “crude to DMSO stock automation system” (1. Waters Autopurify HPLC-MS. 2. Waters Acquity UPLC-MS/CAD/DAD. 3. Porvair Mistral blowdown dryer. 4. Thermo Spinnaker robotic arm. 5. Formulatrix Mantis microfluidic dispenser) B. Laboratory based setup.
      Samples are first injected into the preparative LC-MS and mass-directed fractionation is triggered on the target's product mass, in either positive or negative ionization mode (determined by the ionization profile observed in the previous UHPLC-MS step). The fractions, containing up to 2.4 mL per well, are collected into 96-deep-well plates with a defined number of wells reserved per injection. It is critical that the number of fractions is known to the automation system, so that the scheduler can correctly determine when a collection plate has been fully utilized. Given the application for biopharmaceutical research, the workflow primarily handles compounds with reasonable ionization under positive or negative electrospray ionization conditions and is not suited for compounds requiring specialized MS conditions.
      The purification process is initiated by uploading a comma-separated value (csv) file containing the purification conditions and target list produced by Analytical Studio in the prior scouting workflow. Momentum then parses this file and creates individual text files for each reaction mixture. These text files are deposited one at a time into the MassLynx AutoLynx application to commence the purification of each sample (Fig. 4). Upon finalization of a preparative RP-HPLC injection, the fractions are mixed by undergoing a simulated QC step through the MassLynx software. This fraction mixing was found to be critical as compound concentration is typically highly stratified within the collection vessel. Next, the Spinnaker arm moves the deep-well plate containing the collected fractions into the Acquity UHPLC-MS/CAD system and a new text file is generated. Submission of the latter to AutoLynx on the Acquity system triggers the analysis of the collected fractions. Employing the previously described array of detectors enables simultaneous determination of purity and quantification by charged aerosol detection.[
      • Zhang K.
      • Kurita K.L.
      • Venkatramani C.
      • Russell D.
      Seeking universal detectors for analytical characterizations.
      ,
      • Dixon R.W.
      • Peterson D.S.
      Development and testing of a detection method for liquid chromatography based on aerosol charging.
      ,
      • Almeling S.
      • Ilko D.
      • Holzgrabe U.
      Charged aerosol detection in pharmaceutical analysis.
      ,
      • Ligor M.
      • Studzińska S.
      • Horna A.
      • Buszewski B.
      Corona-charged aerosol detection: an analytical approach.
      ,
      • Reilly J.
      • Everatt B.
      • Aldcroft C.
      Implementation of charged aerosol detection in routine reversed phase liquid chromatography methods.
      ]. The target compounds purified on this platform come from small-molecule drug discovery efforts and are typically suitable for UV or CAD-based purity determination. This combination of a purification process and subsequent QC will continue until all the available wells of a fraction plate have been filled. At this point, the plate is transferred to the Ultravap Mistral and the solvent is removed under a continuous stream of heated nitrogen.
      Fig 4
      Fig 4Workflow overview of automated purification, fraction QC and re-dissolution.
      The Analytical Studio software captures and analyzes LC data generated during preparative HPLC and QC stages of the library purification workflow. It calculates DAD- and CAD-based purity, consequently also enabling purity assessment of non-chromophoric and volatile substances.[
      • Cohen R.D.
      • Liu Y.
      • Gong X.
      Analysis of volatile bases by high-performance liquid chromatography with aerosol-based detection.
      ,
      • Schilling K.
      • Krmar J.
      • Maljurić N.
      • Pawellek R.
      • Protić A.
      • Holzgrabe U.
      Quantitative structure-property relationship modeling of polar analytes lacking UV chromophores to charged aerosol detector response.
      ] In the absence of any detector signal, the respective fraction is flagged by the software and will not be processed further. CAD peak areas are extracted for the desired product peak in each fraction for quantification. Upon review by the operator, this data is exported as a set of .xml files and forwarded to the Momentum scheduler software. Momentum uses a Python script to calculate the volume of DMSO required to generate a 10 mM solution for each fraction which is also written to a separate csv file summarizing the results for the entire plate. The dried fractions are subsequently re-solubilized in DMSO using the Mantis microfluidic dispenser. In order to reduce water absorption by the solvent each plate is lidded following DMSO addition and then stored in a hotel.
      After the automated purification process completes, an operator removes the fraction plates and subsequent pooling of fractions, cherry-picking and transfer into barcoded matrix tubes is performed on a Tecan Freedom EVO liquid-handling platform (Tecan Group AG, Männerdorf, Switzerland). Our internal LIMS system contains the pooling logic, it provides a GUI (graphical user interface) for fraction selection and automatically creates the corresponding transfer lists. A final assessment of the pooled compounds is performed by LC-MS/CAD/DAD to ensure that no degradation occurred during the drying, re-dissolution, or pooling steps as well as to confirm final concentration. The LIMS system facilitates registration of the compounds with internal databases and for updating the Novartis electronic laboratory journal. The 10 mM stock solutions are directly tested in in vitro assays or submitted to the Novartis Compound Hub for further distribution and storage. For assay profiling a dilution series is generated using the Tecan liquid handler according to the project format and transferred into an acoustic dispensing compatible plate. We then utilize an Echo instrument (Beckman Coulter, Brea, CA, USA) to generate the various plate layouts and dilutions required for assays. One consideration for such a purification system is the water content of the DMSO solutions. From an Echo survey of our final compound plates, the water content ranges between 10-20%, which is within the tolerated range of the Echo instrument. While the hygroscopic nature of DMSO results in adsorption of water and consequently a reduction in the concentration of the drug candidates these deviations are significantly lower than the error margins found for typical biochemical assays.[
      • White J.R.
      • Abodeely M.
      • Ahmed S.
      • Debauve G.
      • Johnson E.
      • Meyer D.M.
      • Mozier N.M.
      • Naumer M.
      • Pepe A.
      • Qahwash I.
      • Rocnik E.
      • Smith J.G.
      • Stokes E.SE
      • Talbot J.J.
      • Wong P.Y
      Best practices in bioassay development to support registration of biopharmaceuticals.
      ] Furthermore, the increased accuracy provided by the CAD-based quantitation over gravimetric solution preparation with respect to salt form and residual solvents helps to negate errors resulting from water absorption. In order to further reduce water absorption, the DMSO used in the Mantis liquid dispenser is stored in a capped bottle over molecular sieves and is changed at regular intervals.
      The inclusion of physical automation into this process with the Spinnaker robotic arm results in considerable efficiency gains over a more segmented workflow. Manual purification of a library with 96 members including drying and restocking requires regular human intervention at several stages, thus preventing operation overnight or throughout weekends. By combining semi-preparative purification with immediate QC and quantification, the evaporation of the mobile phase from the deep-well plate while subsequent samples are processed is minimized. This enables direct quantification from a known volume, a prerequisite for CAD-based quantity determination. In the context of the automation setup described here, a full 96-membered library was successfully purified and transformed into 10 mM DMSO stock solutions without human intervention within 42 h.[

      A table with start and stop timepoints can be found in the supporting information.

      ]

      3. Results & discussion

      3.1 Validation

      First, we validated both physical and data workflows to ensure that the Spinnaker robot moves plates and samples successfully between periphery systems and that the required files are created by Momentum and moved between folders without error. A customized nest was designed and produced by Thermo CRS (Thermo CRS Ltd., Burlington, ON, Canada) to enable robotic access into the Acquity Autosampler through the port originally intended for a Waters Sample Organizer. Given the open deck of the AutoPurification platform, no modification was required. However, the software positions on the preparative autosampler deck needed to be reconfigured in MassLynx to utilize only the frontmost row of nest locations because the Spinnaker did not have sufficient reach to access the backmost. The Ultravap Mistral, the Spinnaker and the Mantis are both commonly deployed automation systems and were fully compatible with the ANSI/SLAS microplate standard.
      The CAD-based quantification process was compared to other quantification methodologies and the concentration of the resulting DMSO stock was determined by quantitative nuclear magnetic resonance (qNMR) to assess potential deviation of the target concentration from 10 mM. While qNMR is a highly accurate and universal approach to compound quantitation, it is limited in throughput and requires significant data interpretation. Quantification by CAD or qNMR also enables direct measurement of the compound of interest, removing the effects of residual water, solvents or salt form. We found that normalizing the gradient composition into the CAD to isocratic conditions by addition of a secondary solvent pump provided a more universal signal, which is consistent with the literature.[

      A table listing retention times and the corresponding focused gradient windows can be found in the supporting information.

      ,
      • Zhang K.
      • Kurita K.L.
      • Venkatramani C.
      • Russell D.
      Seeking universal detectors for analytical characterizations.
      ,
      • Crafts C.
      • Bailey B.
      • Acworth I.
      Single calibrant approach for the analysis of unknowns using dual gradient pump and charged aerosol detection.
      ,
      • Górecki T.
      • Lynen F.
      • Szucs R.
      • Sandra P.
      Universal response in liquid chromatography using charged aerosol detection.
      ,
      • Shaodong J.
      • Lee W.J.
      • Ee J.W.
      • Park J.H.
      • Kwon S.W.
      • Lee J.
      Comparison of ultraviolet detection, evaporative light scattering detection and charged aerosol detection methods for liquid-chromatographic determination of anti-diabetic drugs.
      ,
      • Hazotte A.
      • Libong D.
      • Matoga M.
      • Chaminade P.
      Comparison of universal detectors for high-temperature micro liquid chromatography.
      ] The CAD-based quantification was tested on a set of compounds of known concentration derived from a set of internal reference compounds. We have found that, as evidenced elsewhere, the CAD provides an accurate determination of compound concentration.[

      Gamache, P.; Muellner, T.; Eggart, B.; Lovejoy, K.; Acworth, I. Charged aerosol detection – use of the power function and robust calibration practices to achieve the quantitative results. ThermoScientific Technical Note 73299, 2019.

      ,

      Analogous compounds for calibration are typically obtained from drug discovery project teams. In the context of the library utilized for validation of the workflow an analogous firefly luciferase inhibitor was obtained from the Novartis compound archive and a dilutions series was prepared.

      ] According to literature reports, the detector response for CAD is independent from the chemical structure and similar responses should be obtained for different compounds within a comparable molecular weight range.[
      • Dixon R.W.
      • Peterson D.S.
      Development and testing of a detection method for liquid chromatography based on aerosol charging.
      ,
      • Zhang K.
      • Kurita K.L.
      • Venkatramani C.
      • Russell D.
      Seeking universal detectors for analytical characterizations.
      ,
      • Sun P.
      • Wang X.
      • Alquier L.
      • Maryanoff C.A.
      Determination of relative response factors of impurities in paclitaxel with high performance liquid chromatography equipped with ultraviolet and charged aerosol detectors.
      ,
      • Holzgrabe U.
      • Nap C.J.
      • Kunz N.
      • Almeling S.
      Identification and control of impurities in streptomycin sulfate by high-performance liquid chromatography coupled with mass detection and corona charged-aerosol detection.
      ,
      • Gamache P.H.
      • McCarthy R.S.
      • Freeto S.M.
      • Asa D.J.
      • Woodcock M.J.
      • Laws K.
      • Cole R.O.
      HPLC analysis of non-volatile analytes using charged aerosol detection.
      ,
      • Liu G.
      • Zhu B.
      • Ren X.
      • Wang J.
      Universal response method for the quantitative analysis of multi-components in josamycin and midecamycin using liquid chromatography coupled with charged aerosol detector.
      ] We nonetheless chose to prepare serial dilutions of four reference compounds (glyburide, fenofibrate, ketoprofen, and glipizide) across a concentration range and obtain the corresponding calibration functions to demonstrate this. These compounds were each weighed from purchased powder (Sigma Aldrich) and diluted to 10 mM in d6-DMSO. The resulting solutions were then quantified by qNMR (Table 1). The remaining 10 mM solution was diluted by a factor of 2 to avoid over-saturating the CAD detector and a CAD peak area was obtained from the Acquity UHPLC-MS/CAD/DAD instrument. The concentration was subsequently assessed by utilizing a generic calibration curve prepared from a set of 80 proprietary Novartis chemical compounds that represent a wide range of chemical space across various drug discovery projects (Fig. 5). The R2 value for the resulting least-squares regression line was 0.904 and the F-test p-value was <0.01 indicating good linear fit and significant overall results with respect to the data. The resulting slope and intercept (1.02, -0.87) from the calibration curve were input into Equation 1 and used to determine the concentration for unknown samples. Each result was multiplied by a factor of 2 to account for the dilution of the samples for CAD.
      Table 1qNMR quantitation and calibration curves of reference compounds.
      CompoundMolecular weight (g/mol)NMR Shift for Integral (ppm)NMR Proton Area# of ProtonsNMR Conc (mM)NMR Conc (mg/mL)
      Maleic Acid (1mM)1166.27-6.24739160820.990.11
      Maleic Acid (5mM)1166.28-6.253679229224.580.53
      Maleic Acid (10mM)1166.29-6.267440386529.181.06
      Glyburide4947.65-7.6240985399110.104.99
      Glipizide445.58.62-8.5942590251110.504.68
      Fenofibrate360.86.93-6.8882242039210.143.66
      Ketoprofen254.37.95-7.36542481287914.823.77
      Fig 5
      Fig. 5CAD calibration curves for reference compounds.
      CalculatedConcentration(mg/mL)=10log10(CADarea)*1.020.87
      Equation 1: Experimentally derived calibration function
      We compared the resulting concentrations with those generated from a direct integration of the qNMR samples and calculation based on an external standard curve made from maleic acid. The results are shown in Table 2. The quantification by CAD revealed that each of the four compounds were within a 10% mean average deviation of the concentration derived by qNMR (Table 2). When reviewing the corresponding calculated molarities, a mean standard deviation of 0.45 mM was obtained which corresponded to a 4% coefficient of variation between the two quantification methodologies. These results demonstrate that our experimentally derived CAD concentration curve can be used to link a CAD peak area to a concentration for a novel sample and are generally in excellent agreement with qNMR. In the context of biochemical in vitro profiling for early drug discovery, these deviations are well below error margins found for typical TR-FRET-, luminescence- or MS-based assays.[
      • Fay M.P.
      • Sachs M.C.
      • Miura K.
      Measuring precision in bioassays: rethinking assay validation.
      ] These quantitation results demonstrate that the CAD quantification approach is enabling rapid profiling of libraries for SAR expansion while maintaining a high level of rigor to support quantitative data as well as capturing negative SAR.
      Table 2Comparison of qNMR and CAD based concentration determination.
      CompoundMolecular weight (g/mol)CAD AreaCAD Conc (mg/mL)NMR Conc (mM)CAD (mM) Corrected for DilutionError PercentStandard DeviationCoefficient of variation
      Glyburide494.016.062.2910.109.278%0.596%
      Glipizide445.514.742.1010.509.4210%0.768%
      Fenofibrate360.812.741.8110.1410.021%0.081%
      Ketoprofen254.313.711.9514.8215.323%0.362%
      Mean6%0.454%
      For the validation of preparative chromatography, a recovery test was performed on the AutoPurification HPLC system. To this end, a CAD calibration curve was generated using flavone (2-phenyl-4H-1-benzopyran-4-one) at known concentrations and subsequently a known amount of flavone was injected into the HPLC system, fractions were collected using an MS-based trigger and the CAD response of each fraction was measured. Based on the CAD area and the calibration curve, it was determined that >90% of the flavone initially injected into the system was recovered and quantified. A high recovery rate is critical in the context of microscale chemistry and drug discovery and demonstrates that sufficient material can be isolated for subsequent profiling even with a decrease in reaction scale. The primary losses of recovery of sample during the process are during the sample injection process as well as during the collection of fractions, as the flow does not stop when the fraction collection system moves between tubes.
      The communication between the Momentum scheduling software and the LC instruments is accomplished utilizing the AutoLynx program. AutoLynx serves as a file monitoring software which initiates an operation on the LC instrumentation after a correctly formatted file is deposited in a monitored folder. In addition, Momentum monitors the status of the LC instruments through the status file. This file is produced by MassLynx at a given interval and contains information about whether the system is operating or in a state of error as well as information about the current queue of tasks. Additional information for this file workflow can be found in the MassLynx 4.1 interfacing guide.[

      Waters MassLynx 4.1 Interfacing guide, 71500123505/Revision A. Last accessed on February 22nd 2022.

      ] Both the Mistral and the Mantis have an API-based interface with appropriate drivers, so the communication is handled by the Momentum software.

      3.2 Purification of a chemical library

      To assess the utility of the automation platform as well as both physical and data workflows, a chemical library on microscale (5 µmol per reaction) with 96 members was synthesized. The crude reaction mixtures were submitted to the automation platform for preparative purification, quantification, and re-stocking to 10 mM DMSO solutions. A known inhibitor of the firefly luciferase reporter was chosen from the literature for library-based diversification (Fig. 6, compound 1).[
      • Thorne N.
      • Shen M.
      • Lea W.A.
      • Simeonov A.
      • Lovell S.
      • Auld D.S.
      • Inglese J.
      Firefly luciferase in chemical biology: a compendium of inhibitors, mechanistic evaluation of chemotypes, and suggested use as a reporter.
      ] A set of 96 acetic acid derivatives were coupled to methyl 3-aminobenzoate (CAS 4518-10-9) using classic amide coupling conditions in DMA (dimethyl acetamide, 50 µL, 0.1 M) at room temperature (Fig. 6). The reaction was performed in an aluminum Para-dox plate with 96 distinct glass vials and individual stirring bars. Upon completion of the reaction and subsequent filtration, aliquots (2 µL per well) were taken from the crude reaction mixtures by robotic liquid handling to prepare daughter plates for analytical UHPLC-MS analysis (diluted to a final volume of 52 µL per well).
      Fig 6
      Fig. 6Firefly luciferase inhibitor library.
      The Analytical Studio software performed an automated analysis of the LC-MS data and consequently determined which of the 96 reactions would be subjected to preparative HPLC, based on the previously described decision-making capabilities. The software furthermore identified suitable conditions for preparative purification and created a worklist.
      The automated decision-making process was validated through a manual review of the analytical chromatograms by an experienced separation scientist and the generated work order was accepted without modifications. Out of the 96 crude reaction mixtures a total of 92 solutions were submitted to the automated purification process. In the remaining four wells no product was identified by mass spectrometry. In total, 50 crude mixtures were suggested for purification under acidic conditions, while in 42 cases basic conditions were determined to be more suitable. In 39 cases (acidic and basic) a co-elution with the desired product was identified and the gradient duration was extended for increased chromatographic resolution. Prior to injection, the crude reaction mixtures were filtered and diluted in a mixture of acetonitrile, water and DMSO (8:2:1) to a total volume of 300 µL. The automation process was initiated by uploading the work list generated by the Analytical Studio software into the Momentum software. The library purification process including quantification, QC, drying, and restocking was completed within 42 h. Out of the 92 injections, 88 samples resulted in fractions collected by mass-triggering. The remaining four compounds were not present in sufficient quantity to trigger a collection. A total of 81 compounds were isolated in sufficient purity and quantity to enable further in vitro profiling. This corresponds to an overall success rate of 84.4% while successful purification was achieved in 88% of samples where product was detected. The average purity of the purified samples was 93%.[

      Purities were determined by UV and CAD detection. Fractions with purities greater than 95% were captured as 95% pure.

      ] The average yield across the plate was 36% corresponding to an average volume of 215 µL of 10 mM DMSO stock per reaction.
      From the library, five DMSO stock solutions of compounds were picked for qNMR based quantification in order to validate the CAD-based workflow (Table 3). We found that all concentrations were between 8.5 mM to 10.7 mM, resulting in a mean average error from 10 mM concentration of 6%, indicating that the CAD quantification aligns with our previous testing.
      Table 3qNMR control of selected library compounds.
      CompoundStructureqNMR concentration of DMSO stockMean average error from CAD [%]
      EXP001-011-110.2 mM2
      EXP001-029-110.7 mM7
      EXP001-033-110.0 mM0
      EXP001-049-18.5 mM15
      EXP001-070-19.39 mM6.1
      Since introduction of the platform in late 2018, a total of 8759 unique compounds across 308 high-throughput experiments were isolated and registered with a mean purity of 94.1% and median purity of 95.0% (Fig. 7A). The standard deviation of the purity is 2.3% indicating that the samples are generally of high purity. It should be noted that the maximum reported purity for the platform is 95.0%, to account for the limited NMR characterization of the compounds and avoid over interpreting the UV & CAD purity estimates. Therefore, the mean purity reported here is likely an under-estimate. The mean isolated yield was found to be 31.6% and the median isolated yield was found to be 24.3% which is in line with expectations for a high-throughput parallel chemistry workflow. Consequently, compounds obtained through this workflow possess a mean isolated volume of 166.5 µL (median volume is identical) of a 10 mM DMSO solution (Fig. 7B). For most samples this amount of 10 mM stock is sufficient for biochemical or cellular profiling in dose-response format (30 µL), a primary set of physiochemical and ADME assays (10 µL) as well as for submission of the remainder to the Novartis compound archive. Therefore, it is informative to look at the DMSO volumes binned into relative categories as shown in Fig. 8. With <30 µL not being sufficient for complete compound profiling, 30-50 µL being sufficient for a single round of assay profiling and >50 µL being sufficient for assay profiling as well as inclusion of the stock into the compound archive. From the historical data, 78% of the compounds meet this third category, which has meant not only an influx of data for projects, but also an increased enrichment of the compound archive. These key performance indicators demonstrate the usefulness of such an integrated platform for early hit-to-lead discovery, enabling project teams to quickly generate SAR and broadly assess a given hit series with high speed.
      Fig 7
      Fig. 7A. Average compound purity for the platform. B. Average volume of 10 mM DMSO solutions.
      Fig 8
      Fig. 8Binned compound amounts from purification platform.

      4. Outlook

      In its current state, this automation platform enables the purification of up to four libraries per week and consequently has an overall capacity of about 10,000 to 15,000 compounds per year. This throughput matches the output of the semi-automated Novartis micro-scale library workflow today. While the capabilities of future parallel chemistry approaches could saturate the capacity of this automation platform, it offers a significant improvement in throughput over a manual workflow. The automation platform eliminates several manual steps or plate transfers, one dry-down step, and facilitates off-hours equipment operation as well as automated interpretation of chromatograms and subsequent decision making. As future improvements to the platform we envision enabling parallel processing of fraction QC samples with simultaneous purification. We estimate this measure will improve cycle-times by around 30%. The utilization of preparative chromatography columns featuring sub-5 µm particles in combination with column and solvent temperature control will furthermore enable shortening gradient run times by approximately 25%. There is also the opportunity to include additional preparative and analytical LC instruments around the robotic arm to enable further parallelization if the need for increased throughput arises. Furthermore, the combination of RP-HPLC with an orthogonal purification method such as solid-phase extraction (SPE) as part of the overall workflow could enable faster purification methods to be utilized while achieving similar purity profiles.
      Finally, we envision direct communication of LC instruments and the respective acquisition software with the master scheduling software through APIs to replace the file-drop setup and to increase the robustness of the platform to network or file access conflicts. This measure will eliminate waiting times in the scheduling process and remove current limitations such as fraction location pre-assignment.

      5. Conclusion

      We herein disclosed a novel high-throughput robotic platform capable of transforming crude reaction mixtures into 10 mM DMSO stock with high purity, featuring full automation of both physical handling of plates as well as data processing. The platform combines analytical UHPLC-MS with preparative HPLC, mass-triggered fractionation, integrated QC, and quantification as well as drying and re-solubilization. Furthermore, the platform enables pooling and transfer into barcoded matrix tubes for downstream profiling in assays. This platform represents a key component connecting high-throughput parallel synthesis with compound profiling, enabling the generation of high-quality dose-response assay data. It increases the overall efficiency of LC purification processes in drug discovery by using intelligent software with decision making capabilities. Key features include the automation of preparative purification and QC as well as enabling operation overnight and throughout weekends through the removal of manual plate transfers and autonomous decision-making. The scale-down of column size and flowrate facilitated by the reduction in chemistry scale also led to a significant reduction in acetonitrile utilization, helping to reduce waste from parallel synthesis efforts. Furthermore, this automated workflow is fully solution-based and consequently eliminates the need to handle powders or weighing of solids, adding to the robustness and efficiency increase of the platform over manual workflows. The resulting compounds have been applied to multiple internal drug discovery projects and excellent reproducibility was demonstrated when repeating assay results from resynthesized material on traditional scale. The rapid influx of compounds has enabled project teams to quickly build and understand SAR across scaffolds, thus contributing to data driven decision making while expediting drug discovery timelines.

      Author contributions

      The manuscript was written through contributions of all authors. All authors have given approval to the final version of the manuscript.

      Funding sources

      No external funding was received to support this research.

      Declaration of Competing Interest

      The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

      Acknowledgment

      Contributions to developing the workflow by Daniel Gosling and William Ulmer are gratefully acknowledged. Experimental support by Holly Davis is gratefully acknowledged.

      References

        • Schneider G.
        Automating drug discovery.
        Nat Rev Drug Discov. 2018; 17: 97-113
        • Peplow M.
        The robo-chemist.
        Nature. 2014; 512: 20-22
        • Guetzoyan L
        • Ingham R.J.
        • Nikbin N.
        • Rossignol J.
        • Wolling M.
        • Baumert M.
        • Burgess-Brown N.A.
        • StrainDamerell C.M.
        • Shrestha L.
        • Brennan P.E.
        • Fedorov O.
        • Knapp S.
        • Ley S.V.
        Machine-assisted synthesis of modulators of the histone reader BRD9 using flow methods of chemistry and frontal affinity chromatography.
        Med Chem Commun. 2014; 5: 540-546
        • McMullen J.P.
        • Jensen K.F.
        Integrated microreactors for reaction automation: new approaches to reaction development.
        Annu Rev Anal Chem. 2010; 3: 19-42
        • Elvira K.S.
        • Casedevall i Solvas X.
        • Wootton R.C.R.
        • deMello A.J.
        The past, present and potential for microfluidic reactor technology in chemical synthesis.
        Nat Chem. 2013; 5: 905-915
        • Hwang Y.-J.
        • Coley C.W.
        • Abolhasani M.
        • Marzinzik A.L.
        • Koch G.
        • Spanka C.
        • Lehmann H.
        • Jensen K.F.
        Segmented flow platform for on-demand medicinal chemistry and compound synthesis: efficient reaction screening in oscillating droplets.
        Chem Commun. 2017; 53: 6649-6652
        • Michael Werner M.
        • Kuratli C.
        • Martin R.E.
        • Hochstrasser R.
        • Wechsler D.
        • Enderle T.
        • Alanine A.I.
        • Vogel H
        Seamless integration of dose-response screening and flow chemistry: efficient generation of structure–activity relationship data of β-Secretase (BACE1) inhibitors.
        Angew Chem Int Ed. 2014; 53: 1704-1708
        • Czechtizky W.
        • Dedio J.
        • Desai B.
        • Dixon K.
        • Farrant E.
        • Feng Q.
        • Morgan T.
        • Parry D.M.
        • Ramjee M.K.
        • Selway C.N.
        • Schmidt T.
        • Tarver G.J.
        • Wright A.G.
        Integrated synthesis and testing of substituted xanthine based DPP4 inhibitors: application to drug discovery.
        ACS Med Chem Lett. 2013; 4: 768-772
        • Chow S.
        • Liver S.
        • Nelson A.
        Streamlining bioactive molecular discovery through integration and automation.
        Nat Rev Chem. 2018; 2: 174-183
        • Baranczak A.
        • Tu N.P.
        • Marjanovic J.
        • Searle P.A.
        • Vasudevan A.
        • Djuric S.W.
        Integrated platform for expedited synthesis-purification-testing of small molecule libraries.
        ACS Med Chem Lett. 2017; 8: 461-465
        • Buitrago Santanilla A.
        • Regalado E.L.
        • Pereira T.
        • Shevlin M.
        • Bateman K.
        • Campeau L.-C.
        • Schneeweis J.
        • Berritt S.
        • Shi Z.-C.
        • Nantermet P.
        • Liu Y.
        • Helmy R.
        • Welch C.J.
        • Vachal P.
        • Davies I.W.
        • Cernak T.
        • Dreher S.D.
        Nanomole-scale high-throughput chemistry for the synthesis of complex molecules.
        Science. 2015; 347: 49-53
        • Mitchell M.C.
        • Spikmans V.
        • de Mello A.J.
        Microchip-based synthesis and analysis: control of multicomponent reaction products and intermediates.
        Analyst. 2001; 126: 24-27
        • Garcia-Egido E.
        • Spikmans V.
        • Wong S.Y.F.
        • Warrington B.H.
        Synthesis and analysis of combinatorial libraries performed in an automated micro reactor system.
        Lab Chip. 2003; 3: 73-76
        • Shaabani S.
        • Xu R.
        • Ahmadianmoghaddam M
        • Gao L.
        • Stahorsky M.
        • Olechno J.
        • Ellson R.
        • Kossenjans M.
        • Helan V.
        • Dömling A.
        Automated and accelerated synthesis of indole derivatives on a nanoscale.
        Green Chem. 2019; 21: 225-232
        • Gesmundo N.J.
        • Sauvagnat B.
        • Curran P.J.
        • Richards M.P.
        • Andrews C.L.
        • Dandliker P.J.
        • Cernak T.
        Nanoscale synthesis and affinity ranking.
        Nature. 2018; 557: 228-232
        • Osipyan A.
        • Shaabani S.
        • Warmerdam R.
        • Shishkina S.V.
        • Boltz H.
        • Dömling A.
        Automated, accelerated nanoscale synthesis of iminopyrrolidines.
        Angew Chem Int Ed. 2020; 59: 1-7
        • Weller H.N.
        • Nirschl D.S.
        • Paulson J.L.
        • Hoffman S.L.
        • Bullock W.H.
        Addressing the medicinal chemistry bottleneck: a lean approach to centralized purification.
        ACS Comb Sci. 2012; 14: 520-526
        • Liu M.
        • Kuanchang C.
        • Christian D.
        • Fatima T.
        • Pisarnitski N.
        • Streckfuss E.
        • Zhang C.
        • Xia L.
        • Borges S.
        • Shi Z.
        • Vachal P.
        • Tata J.
        • Athanasopoulos J.
        High-throughput purification platform in support of drug discovery.
        ACS Comb Sci. 2012; 14: 51-59
        • Wang Y.
        • Shaabani S.
        • Ahmadianmoghaddam M.
        • Gao L.
        • Xu R.
        • Kurpiewska K.
        • Kalinowska-Tluscik J.
        • Olechno J.
        • Ellson R.
        • Kossenjans M.
        • Helan V.
        • Groves M.
        • Dömling A.
        Acoustic droplet ejection enabled automated reaction scouting.
        ACS Cent Sci. 2019; 5: 451-457
        • Desai B.
        • Dixon K.
        • Farrant E.
        • Feng Q.
        • Gibson K.R.
        • van Hoorn W.P.
        • Mills J.
        • Morgan T.
        • Parry D.M.
        • Ramjee M.K.
        • Selway C.N.
        • Tarver G.J.
        • Whitlock G.
        • Wright A.G.
        Rapid discovery of a novel series of AbI kinase inhibitors by application of an integrated microfluidic synthesis and screening platform.
        J Med Chem. 2013; 56: 3033-3047
        • Gawehn E.
        • Hiss J.A.
        • Schneider G.
        Deep learning in drug discovery.
        Mol Inf. 2016; 35: 3-14
        • Gao K.
        • Shaabani S.
        • Xu R.
        • Zarganes-Tzitzikas T.
        • Gao L.
        • Ahmadianmoghaddam M.
        • Groves M.R.
        • Dömling A.
        Nanoscale, automated, high throughput synthesis and screening for the accelerated discovery of protein modifiers.
        RSC Med Chem. 2012; 12: 809-818
        • O'Connell T.N.
        • Ramsay J.
        • Rieth J.R.
        • Shapiro M.J.
        • Stroh J.G.
        Solution-based indirect affinity selection mass spectroscopy—a general tool for high-throughput screening of pharmaceutical compound libraries.
        Anal Chem. 2014; 86: 7413-7420
        • Gabriel J.
        • Höfner G.
        • Wanner K.T.
        A library screening strategy combining the concepts of MS binding assays and affinity selection mass spectrometry.
        Front Chem. 2019; 7 (665): 1-18
        • Qin S.
        • Meng M.
        • Yang D.
        • Bai W.
        • Lu Y.
        • Peng Y.
        • Song G.
        • Wu Y.
        • Zhou Q.
        • Zhao S.
        • Huang X.
        • McCorvy J.D.
        • Cai X.
        • Dai A.
        • Roth B.L.
        • Hanson M.A.
        • Liu Z.-J.
        • Wang M.-W.
        • Stevens R.C.
        • Shui W.
        High-throughput identification of G protein-coupled receptor modulators through affinity mass spectrometry screening.
        Chem Sci. 2018; 9: 3192-3199
        • Blakskjaer P.
        • Heitner T.
        • Hansen N.J.V.
        Fidelity by design: yoctoreactor and binder trap enrichment for small-molecule DNA-encoded libraries and drug discovery.
        Curr Opin Chem Biol. 2015; 26: 62-71
        • Satz A.L.
        • Kuai L.
        • Peng X.
        Selections and screenings of DNA-encoded chemical libraries against enzymes and cellular targets.
        Bioorg Med Chem Lett. 2021; 39127851
        • Reddavide F.V.
        • Cui M.
        • Lin W.
        • Fu N.
        • Heiden S.
        • Andrade H.
        • Thompson M.
        • Zhang Y.
        Second generation DNA-encoded dynamic combinatorial chemical libraries.
        Chem Commun. 2019; 55: 3753-3756
        • Dykstra K.D.
        • Streckfuss E.
        • Liu M.
        • Liu J.
        • Yu Y.
        • Wang M.
        • Kozlowski J.A.
        • Myers R.W.
        • Buevich A.V.
        • Maletic M.M.
        • Vachal P.
        • Krska S.W.
        Synthesis of HDAC inhibitor libraries via mircoscale workflow.
        ACS Med Chem Lett. 2021; 12: 337-342
        • Hochlowski J.E.
        • Searle P.A.
        • Tu N.P.
        • Pan J.Y.
        • Spanton S.G.
        • Djuric S.W.
        An integrated synthesis-purification system to accelerate the generation of compounds in pharmaceutical discovery.
        J Flow Chem. 2011; 2: 56-61
        • Sutherland J.D.
        • Tu N.P.
        • Nemcek T.A.
        • Searle P.A.
        • Hochlowski J.E.
        • Djuric S.W.
        • Pan J.Y.
        An automated synthesis-purification-sample-management platform for the accelerated generation of pharmaceutical candidates.
        J Lab Autom. 2014; 19: 176-182
        • Lall M.S.
        • Bassyouni A.
        • Bradow J.
        • Brown M.
        • Bundesmann M.
        • Chen J.
        • Ciszewski G.
        • Hagen A.E.
        • Hyek D.
        • Jenkinson S.
        • Liu B.
        • Obach R.S.
        • Pan S.
        • Reilly U.
        • Sach N.
        • Smaltz D.J.
        • Spracklin D.K.
        • Starr J.
        • Wagenaar M.
        • Walker G.S.
        Late-stage lead diversification coupled with quantitative nuclear magnetic resonance spectroscopy to identify new structure–activity relationship vectors at nanomole-scale synthesis: application to loratadine, a human histamine H1 receptor inverse agonist.
        J Med Chem. 2020; 63: 7268-7292
        • Chow S.
        • Liver S.
        • Nelson A.
        Streamlining bioactive molecular discovery through integration and automation.
        Nat Rev Chem. 2018; 2: 174-183
        • Barhate C.L.
        • Donnell A.F.
        • Davies M.
        • Li L.
        • Zhang Y.
        • Yang F.
        • Black R.
        • Zipp G.
        • Zhang Y.
        • Cavallaro C.
        • Priestly E.S.
        • Weller H.N.
        Microscale purification in support of high-throughput medicinal chemistry.
        Chem Commun. 2021; 57: 11037-11040
        • Ellson R.
        Picoliter: enabling precise transfer of nanoliter and picoliter volumes.
        Drug Discov Today. 2002; 7: S32-S34
        • Dawes T.D.
        • Turincio R.
        • Jones S.W.
        • Rodriguez R.A.
        • Gadiagellan D.
        • Thana P.
        • Clark K.R.
        • Gustafson A.E.
        • Orren L.
        • Liimatta M.
        Compound transfer by acoustic droplet ejection promotes quality and efficiency in ultra-high-throughput screening campaigns.
        J Lab Autom. 2016; 21: 64-75
        • Wang J.
        • Holloway T.
        • Lisova K.
        • van Dam R.M.
        Green and efficient synthesis of the radiopharmaceutical [18F]FDOPA using a microdroplet reactor.
        React Chem Eng. 2020; 5: 320-329
      1. Aubin, A. J.; Cleary, R. Analytical HPLC to preparative HPLC: scale-up techniques using a natural product extract. Waters Application Note 720003120 2019.

      2. A table listing retention times and the corresponding focused gradient windows can be found in the supporting information.

        • Zhang K.
        • Kurita K.L.
        • Venkatramani C.
        • Russell D.
        Seeking universal detectors for analytical characterizations.
        J Pharm Biomed Anal. 2019; 162: 192-204
        • Dixon R.W.
        • Peterson D.S.
        Development and testing of a detection method for liquid chromatography based on aerosol charging.
        Anal Chem. 2002; 74: 2930-2937
        • Almeling S.
        • Ilko D.
        • Holzgrabe U.
        Charged aerosol detection in pharmaceutical analysis.
        J Pharm Biomed Anal. 2012; 69: 50-63
        • Ligor M.
        • Studzińska S.
        • Horna A.
        • Buszewski B.
        Corona-charged aerosol detection: an analytical approach.
        Crit Rev Anal Chem. 2013; 43: 64-78
        • Reilly J.
        • Everatt B.
        • Aldcroft C.
        Implementation of charged aerosol detection in routine reversed phase liquid chromatography methods.
        J Liq Chromatogr Relat. 2008; 31: 3132-3142
        • Cohen R.D.
        • Liu Y.
        • Gong X.
        Analysis of volatile bases by high-performance liquid chromatography with aerosol-based detection.
        J Chromatogr A. 2012; 1229: 172-179
        • Schilling K.
        • Krmar J.
        • Maljurić N.
        • Pawellek R.
        • Protić A.
        • Holzgrabe U.
        Quantitative structure-property relationship modeling of polar analytes lacking UV chromophores to charged aerosol detector response.
        Anal Bioanal Chem. 2019; 411: 2945-2959
        • White J.R.
        • Abodeely M.
        • Ahmed S.
        • Debauve G.
        • Johnson E.
        • Meyer D.M.
        • Mozier N.M.
        • Naumer M.
        • Pepe A.
        • Qahwash I.
        • Rocnik E.
        • Smith J.G.
        • Stokes E.SE
        • Talbot J.J.
        • Wong P.Y
        Best practices in bioassay development to support registration of biopharmaceuticals.
        BioTechniques. 2019; 67: 126-137
      3. A table with start and stop timepoints can be found in the supporting information.

        • Crafts C.
        • Bailey B.
        • Acworth I.
        Single calibrant approach for the analysis of unknowns using dual gradient pump and charged aerosol detection.
        Appl Noteb. 2010; 0: 29
        • Górecki T.
        • Lynen F.
        • Szucs R.
        • Sandra P.
        Universal response in liquid chromatography using charged aerosol detection.
        Anal Chem. 2006; 78: 3186-3192
        • Shaodong J.
        • Lee W.J.
        • Ee J.W.
        • Park J.H.
        • Kwon S.W.
        • Lee J.
        Comparison of ultraviolet detection, evaporative light scattering detection and charged aerosol detection methods for liquid-chromatographic determination of anti-diabetic drugs.
        J Pharm Biomed Anal. 2010; 51: 973-978
        • Hazotte A.
        • Libong D.
        • Matoga M.
        • Chaminade P.
        Comparison of universal detectors for high-temperature micro liquid chromatography.
        J Chromatogr A. 2007; 1170: 52-61
      4. Gamache, P.; Muellner, T.; Eggart, B.; Lovejoy, K.; Acworth, I. Charged aerosol detection – use of the power function and robust calibration practices to achieve the quantitative results. ThermoScientific Technical Note 73299, 2019.

      5. Analogous compounds for calibration are typically obtained from drug discovery project teams. In the context of the library utilized for validation of the workflow an analogous firefly luciferase inhibitor was obtained from the Novartis compound archive and a dilutions series was prepared.

        • Sun P.
        • Wang X.
        • Alquier L.
        • Maryanoff C.A.
        Determination of relative response factors of impurities in paclitaxel with high performance liquid chromatography equipped with ultraviolet and charged aerosol detectors.
        J Chromatogr A. 2008; 1177: 87-91
        • Holzgrabe U.
        • Nap C.J.
        • Kunz N.
        • Almeling S.
        Identification and control of impurities in streptomycin sulfate by high-performance liquid chromatography coupled with mass detection and corona charged-aerosol detection.
        J Pharm Biomed Anal. 2011; 56: 271-279
        • Gamache P.H.
        • McCarthy R.S.
        • Freeto S.M.
        • Asa D.J.
        • Woodcock M.J.
        • Laws K.
        • Cole R.O.
        HPLC analysis of non-volatile analytes using charged aerosol detection.
        LC–GC Eur. 2005; 18: 345-354
        • Liu G.
        • Zhu B.
        • Ren X.
        • Wang J.
        Universal response method for the quantitative analysis of multi-components in josamycin and midecamycin using liquid chromatography coupled with charged aerosol detector.
        J Pharm Biomed Anal. 2021; 192113679
        • Fay M.P.
        • Sachs M.C.
        • Miura K.
        Measuring precision in bioassays: rethinking assay validation.
        Stat Med. 2018; 37: 519-529
      6. Waters MassLynx 4.1 Interfacing guide, 71500123505/Revision A. Last accessed on February 22nd 2022.

        • Thorne N.
        • Shen M.
        • Lea W.A.
        • Simeonov A.
        • Lovell S.
        • Auld D.S.
        • Inglese J.
        Firefly luciferase in chemical biology: a compendium of inhibitors, mechanistic evaluation of chemotypes, and suggested use as a reporter.
        Chem Biol. 2012; 19: 1060-1072
      7. Purities were determined by UV and CAD detection. Fractions with purities greater than 95% were captured as 95% pure.