combinatorial chemistry, part b

Preface
Combinatorial chemistry has matured from a field where efforts initially
focused on peptide-based research to become an indispensable research tool
for molecular recognition, chemical-property optimization, and drug discovery.
Originally used as a method to primarily generate large numbers of molecules,
combinatorial chemistry has been significantly influenced and integrated with
other important fields such as medicinal chemistry, analytical chemistry, syn-
thetic chemistry, robotics, and computational chemistry.
Even though the initial focus of attention was providing larger numbers of
molecules with a ‘‘diversity’’ goal in mind, other factors came into play
depending upon the problem scientists were trying to solve, such as bioactivity,
solubility, permeability properties, PK, ADME, toxicity, and patentability.
One can think of combinatorial chemistry and compound screening as an
iterative Darwinian process of divergence and selection. Particularly in drug
discovery, where time is a critical factor to success, combinatorial chemistry
offers the means to test more molecule hypotheses in parallel.
We will always be limited to a finite number of molecules that we can
economically synthesize and evaluate. Even with all the advances in automa-
tion technologies, combinatorial chemistry, and higher-throughput screens that
improve our ability to rapidly confirm or disprove hypotheses, the synthesis
and screening cycle remains the rate-determining process. Fortunately, we
continue to make great strides forward in the quality and refinement of pre-
dictive algorithms and in the breadth of the training sets amassed to aid in the
drug discovery/compound optimization iterative process.
Anyone who has optimized chemical reactions for combinatorial libraries
or process chemistry knows first hand how much experimentation is required to
identify optimal conditions. Chemical feasibility is at the heart of small mol-
ecule discovery and chemotype prioritization since it essentially defines what
can and cannot be analoged (i.e., analogability). Although analogability is not
the only driving factor, quite often it is overlooked. For example, when com-
mercially-available compounds or complex natural products are screened, the

leads generated are often dropped because of the difficulty to rapidly analog
them in the lead optimization stage.
The desirability of a chemotype is a function of drug-likeness, potency,
novelty, and analogability. A particularly attractive feature of combinatorial
chemistry is that when desirable properties are identified, they can often be
xiii
optimized through second-generation libraries following optimized synthetic
protocols. If this process of exploring truly synthetically accessible chemical
spaces could be automated, then it would open up the exciting possibility of
modeling the iterative synthesis and screening cycle.
Predicting, or even just mapping, synthetic feasibility is a sleeping giant;
few people are looking into it, and the ramifications of a breakthrough would
be revolutionary for both chemistry and drug discovery. In-roads to predicting
(or even just mapping) chemical feasibility have the potential to have as large
an impact on drug discovery as computational models of bioavailability and
drugability. These are important questions where scientists are now starting to
generate a large-enough body of information on high-throughput synthetic
chemistry to begin to more globally understand what is cost-effectively pos-
sible. Within the biopharmaceutical industry, significant investments in new
technologies have been made in molecular biology, genomics, and proteomics.
However, with the exception of combinatorial chemistry, relatively little has
been done to advance the fundamental nature of chemistry in drug discovery
from a conceptual perspective.
Now, after having gone through the molecule-generating period where
research institutions have a large historical compound collection and the pro-
liferation of combinatorial chemistry services, the trend is now after making
more targeted-oriented molecular entities also known as ‘‘focused libraries.’’
An important emerging question is: How can one most effectively make the
best possible ‘‘focused libraries’’ to answer very specific research questions,
given all the possible molecules one could theoretically synthesize?

The first installment in this series (Volume 267, 1996) mostly covered
peptide and peptidomimetic based research with just a few examples of small
molecule libraries. In this volume we have compiled cutting-edge research in
combinatorial chemistry, including divergent areas such as novel analytical
techniques, microwave-assisted synthesis, novel linkers, and synthetic ap-
proaches in both solid-phase and polymer-assisted synthesis of peptides, small
molecules, and heterocyclic systems, as well as the application of these tech-
nologies to optimize molecular properties of scientific and commercial interest.
Guillermo A. Morales
Barry A. Bunin
xiv preface
METHODS IN ENZYMOLOGY
EDITORS-IN-CHIEF
John N. Abelson Melvin I. Simon
DIVISION OF BIOLOGY
CALIFORNIA INSTITUTE OF TECHNOLOGY
PASADENA, CALIFORNIA
FOUNDING EDITORS
Sidney P. Colowick and Nathan O. Kaplan
Contributors to Volume 369
Article numbers are in parentheses and following the names of contributors.
Affiliations listed are current.
Fernando Albericio (2), University
of Barcelona, Barcelona Biomedical
Research Institute, Barcelona Science
Park, Josep Samitier 1, Barcelona,
08028, Spain
Alessandra Bartolozzi (19), Surface
Logix, Inc., 50 Soldiers Field Place,
Brighton, Massachusetts, 02135

Hugues Bienayme
´
(24), Chrysalon Mo-
lecular Research, IRC, 11 Albert Einstein
Avenue, Villeurbannem, 69100, France
Sylvie E. Blondelle (18), Torrey Pines
Institute for Molecular Studies, 3550
General Atomics Court, San Diego,
California, 92121
Ce
´
sar Boggiano (18), Torrey Pines
Institute for Molecular Studies, 3550
General Atomics Court, San Diego,
California, 92121
Stefan Bra
¨
se (7), Institut fu
¨
r Organische
Chemie, Universita
¨
t Karlsruhe (TH),
Fritz-Haber-Weg 6, Karlsruhe, D-76131,
Germany
Andrew M. Bray (3), Mimotopes Pty
Ltd., 11 Duerdin Street, Clayton, Vic-
toria, 3168, Australia
Wolfgang K D. Brill (23), Discovery
Research Oncology, Pharmacia Italy

S.p.A, Viale Pasteur 10, Nerviano (MI),
I-20014, Italy
Max Broadhurst (14), Alchemia Pty Ltd.,
Eight Mile Plains, Queensland 4113, Aus-
tralia
Balan Chenera (24), Amgen Inc., Depart-
ment of Small Molecule Drug Discovery,
One Amgen Center Drive, Thousand
Oaks, California, 91320
James W. Christensen (5), Advanced
ChemTech Inc., 5609 Fern Valley Road,
Louisville, Kentucky, 40228
Andrew P. Combs (12), Incyte Corpo-
ration,Wilmington, Delaware, 19880-0500
Scott M. Cowell (16), Department of
Chemistry, University of Arizona,
Tucson, Arizona, 85721
Stefan Dahmen (7), Institut fur Orga-
nische Chemie, RWTH Aachen, Pirlet-
Str. 1, Aachen, 52074, Germany
Ninh Doan (17), Division of Hematology
and Oncology, Department of Internal
Medicine, UC Davis Cancer Center, Uni-
versity of California Davis, Sacramento,
California, 95817
Roland E. Dolle (8), Senior Director of
Chemistry, Department of Chemistry,
Adolor Corporation, 700 Pennsylvania
Drive, Exton, Pennsylvania, 19345
Nicholas Drinnan (14), Alchemia Pty

Ltd., Eight Mile Plains, Queensland
4113, Australia
Amanda M. Enstrom (17), Division of
Hematology and Oncology, Department
of Internal Medicine, UC Davis Cancer
Center, University of California Davis,
Sacramento, California, 95817
ix
Liling Fang (1), ChemRx Division, Dis-
covery Partners International, 385 Oyster
Point Boulevard, Suite 1, South San
Francisco, California, 94080
Eduard R. Felder (23), Discovery Re-
search Oncology, Pharmacia Italy
S.p.A., Viale Pasteur 10, Nerviano
(MI), I-20014, Italy
A
´
rpa
´
d Furka (5), Eo
¨
tvo
¨
s Lora
´
nd Univer-
sity, Department of Organic Chemistry,
P.O. Box 32, Budapest 112, H-1518,
Hungary

A. Ganesan (22), University of Southamp-
ton, Department of Chemistry, Highfield,
Southampton, SO17 1BJ, United Kingdom
J. Gabriel Garcia (20), 4SC AG, Am
Klopferspitz 19A, 82152, Martinsried,
Germany
Brian Glass (13), Incyte Corporation,
Wilmington, Delaware, 19880-0500
Matthias Grathwohl (14), Alchemia Pty
Ltd., Eight Mile Plains, Queenland 4113,
Australia
Michael J. Grogan (19), Surface Logix,
Inc., 50 Soldiers Field Place, Brighton,
Massachusetts, 02135
Xuyuan Gu (16), Department of
Chemistry, University of Arizona,
Tuscon, Arizona, 85721
Eric Healy (5), Advanced ChemTech Inc.,
5609 Fern Valley Road, Louisville,
Kentucky, 40228
Timothy F. Herpin (4), Rho
ˆ
ne-Poulenec
Rorer, 500 Arcola Road, Collegeville,
Pennsylvania, 19426
Cornelia E. Hoesl (25), Torrey Pines In-
stitute, Room 2-136, 3550 General Atom-
ics Court, San Diego, California, 92121
Christopher P. Holmes (9), Affymax Inc.,
4001 Miranda Avenue, Palo Alto,

California, 94304
Richard Houghten (25), Torrey Pines In-
stitute for Molecular Studies, 3550 Gen-
eral Atomics Court, Room 2-136, San
Diego, California, 92121
Victor J. Hruby (16), Department of
Chemistry, University of Arizona,
Tucson, Arizona, 85721
Christopher Hulme (24), Amgen Inc., De-
partment of Small Molecule Drug Discov-
ery, One Amgen Center Drive, 29-1-B,
Thousand Oaks, California, 91320
Sharon A. Jackson (12), Aventis Pharma-
ceuticals, 202-206, Bridgewater, New
Jersey, 08807-0800
Ian W. James (3), Mimotopes Pty Ltd., 11
Duerdin Street, Clayton, Victoria, 3168,
Australia
Wyeth Jones (24), Amgen Inc., Depart-
ment of Small Molecule Drug Discovery,
One Amgen Center Drive, 29-1-B, Thou-
sand Oaks, California, 91320
Patrick Jouin (10), CNRS UPR 9023,
CCIPE, 141, rue de la Cardonille, Mont-
pellier Cedex 05, 34094, France
C. Oliver Kappe (11), Institute of Chemis-
try, Karl-Franzens-University Graz,
Heinrichstrasse 28, Graz, A-8010, Austria
Steven A. Kates (19), Surface Logix, Inc.,
50 Soldiers Field Place, Brighton, Massa-

chusetts, 02135
Viktor Krchn
ˇ
a
´
k (6), Torviq, 3251 West
Lambert Lane, Tuscon, Arizona, 85742
Kit S. Lam (15, 17), Division of Hematol-
ogy and Oncology, Department of In-
ternal Medicine, UC Davis Cancer
Center, University of California Davis,
Sacramento, California, 95817
Alan L. Lehman (17), Division of Hema-
tology and Oncology, Department of In-
ternal Medicine, UC Davis Cancer
Center, University of California Davis,
Sacramento, California, 95817
x contributors to volume 369
Ruiwu Liu (15, 17), Division of Hematol-
ogy and Oncology, Department of In-
ternal Medicine, UC Davis Cancer
Center, University of California Davis,
Sacramento, California, 95817
Matthias Lormann (7), Kekule
´
-Institut fu
¨
r
Organische Chemie und Biochemie der
Rheinischen, Friedrich Wilhelms Univer-

sita
¨
t Bonn, Gerhard-Domagk-Strasse 1,
Bonn, D-53121, Germany
Jan Marik (15), Division of Hematology
and Oncology, Department of Internal
Medicine, UC Davis Cancer Center, Uni-
versity of California Davis, Sacramento,
California, 95817
Katia Martina (23), Discovery Research
Oncology, Pharmacia Italy S.p.A., Viale
Pasteur 10, Nerviano (MI), I-20014, Italy
Joeseph Maxwell (17), Division of Hema-
tology and Oncology, Department of In-
ternal Medicine, UC Davis Cancer
Center, University of California Davis,
Sacramento, California, 95817
Wim Meutermans (14), Alchemia Pty Ltd.,
3 Hi-Tech Court, Brisbane Technology
Park, Eight Mile Plains, QLD 4113, Aus-
tralia
George C. Morton (4), Rho
ˆ
ne-Poulenc
Rorer, 500 Arcola Road, Collegeville,
Pennsylvania, 19426
Adel Nefzi (25), Torrey Pines Institute for
Molecular Studies, 3550 General Atomics
Court, San Diego, California, 92121
Thomas Nixey (24), Amgen Inc., Depart-

ment of Small Molecule Drug Discovery,
One Amgen Center Drive, 29-1-B, Thou-
sand Oaks, California, 91320
John M. Ostresh (25), Torrey Pines Insti-
tute, Room 2-136, 3550 General Atomics
Court, San Diego, California 92121
Vitecek Pade
ˇ
ra (6), Torvic, 3251 W Lam-
bert Lane, Tucson, Arizona, 84742
E.R. Palmacci (13), 77 Massachusetts
Avenue, T18-209, Cambridge, Massachu-
setts, 02139
Yijun Pan (9), Affymax Inc., 4001 Mi-
randa Avenue, Palo Alto, California,
94304
Jack G. Parsons (3), Mimotopes Pty Ltd.,
11 Duerdin Street, Clayton, Victoria,
3168, Australia
Robert Pascal (10), UMR 5073, Univer-
site
´
de Montpellier 2, CC017, place
Euge
`
ne Bataillon, Montpellier Cedex 05,
F-34094, France
Clemencia Pinilla (18), Torrey Pines In-
stitute for Molecular Studies and Mixture
Sciences, Inc., 3550 General Atomics

Court, San Diego, California, 92121
Obadiah J. Plante (13), Massachusetts
Institute of Technology, Department of
Chemistry, 77 Massachusetts Avenue,
Cambridge, Massachusetts, 02139-4307
Gregory Qushair (2), University
of Barcelona, Barcelona Biomedical
Research Institute, Barcelona Science
Park, Josep Samitier 1, Barcelona,
08028, Spain
Jorg Rademann (21), Eberhard-Karls-Uni-
versity, Tu
¨
bingen, Institute of Organic
Chemistry, Auf der Morgenstelle 18, Tu
¨
-
bingen, 72076, Germany
Joseph M. Salvino (8), Director of Com-
binational Chemistry, Adolor Corpor-
ation, 700 Pennsylvania Drive, Exton,
Pennsylvania, 19345
Peter H. Seeberger (13), Laboratorium
fuer Organische Chemie, HCI F 315,
Wolfgang-Pauli-Str. 10, ETH-Hoengger-
berg, CH-8093 Zu
¨
rich, Switzerland
Craig S. Sheehan (3), Mimotopes Pty
Ltd., 11 Duerdin Street, Clayton, Vic-

toria, 3168, Australia
contributors to volume 369 xi
Adrian L. Smith (24), Amgen Inc., Depart-
ment of Small Molecule Drug Discovery,
One Amgen Center Drive, Thousand
Oaks, California, 91320
Re
´
gine Sola (10), UMR 5076, Ecole
Nationale Supe
´
rieure de Chimie de
Montpellier, 8, rue Delaware l’Ecole
Normale, Montpellier Cedex 05, F-
34296, France
Aimin Song (17), University of California,
UC Davis Cancer Center, Division of
Hematology and Oncology, 4501 X
Street, Sacramento, California, 95817
Alexander Stadler (11), Institute of
Chemistry, Karl-Franzens-University
Graz, Heinrichstrasse 28, Graz, A-8010,
Austria
Paul Tempest (24), Amgen Inc., Depart-
ment of Small Molecule Drug Discovery,
One Amgen Center Drive, 29-1-B, Thou-
sand Oaks, California, 91320
David Tumelty (9), Affymax, Inc.,
4001 Miranda Avenue, Palo Alto,
California, 94304

Josef Vagner (16), Department of Chem-
istry, University of Arizona, Tuscon, Ari-
zona, 85741
Jesus Vazquez (2), University of Barce-
lona, Barcelona Biomedical Research
Institute, Barcelona Science Park, Josep
Samitier 1, Barcelona, 08028, Spain
Michael L. West (14), Alchemia Pty Ltd.,
Eight Mile Plains, Queensland 4113,
Australia
Zemin Wu (3), Mimotopes Pty Ltd., 11
Duerdin Street, Clayton, Victoria, 3168,
Australia
Bing Yan (1), ChemRx Division, Discovery
Partners International, 385 Oyster Point,
Boulevard, Suite 1, South San Francisco,
California, 94080
Yongping Yu (25), Torrey Pines Institute,
Room 2-136, 3550 General Atomics
Court, San Diego, California, 92121
Florencio Zaragoza (26), Medicinal
Chemistry, Novo Nordisk A/S, Novo Nor-
disk Park, Malov, 2760, Denmark
Jiang Zhao (1), ChemRx Division, Discov-
ery Partners International, 385 Oyster
Point Boulevard, Suite 1, South San
Francisco, California, 94080
xii contributors to volume 369
[1] High-Throughput Parallel LC/UV/MS Analysis of
Combinatorial Libraries

By Liling Fang,Jiang Zhao,andBing Yan
Introduction
Combinatorial chemistry and high-throughput organic synthesis allow
the preparation of a large number of diverse compounds in a relative short
period of time in order to accelerate discovery efforts in the pharmaceut-
ical and other industries. A library can comprise hundreds to thousands
of compounds with the need to rapidly analyze those compounds for their
identity and purity. Different compound separation and mass spectrometry
(MS) techniques have been applied for the characterization of combinator-
ial libraries. These include separation techniques such as liquid chromatog-
raphy (LC) and capillary electrophoresis and different ionization methods
and mass analyzers.
1–3
LC/MS
*
is the most popular technique used in com-
binatorial library analysis because it combines separation, molecular
weight determination, and relative purity evaluation in a single sample in-
jection. However, the throughput of conventional LC/MS could not meet
the need to analyze every member in a large combinatorial library in a
timely fashion.
Higher-throughput analysis was achieved by utilizing shorter columns
at higher flow rates.
4
Supercritical fluid chromatography (SFC)/MS has
1
A. Hauser-Fang and P. Vouros, ‘‘Analytical Techniques in Combinatorial Chemistry’’
(M. E. Swartz, ed.). Marcel Dekker, New York, 2000.
2
B. Yan, ‘‘Analytical Methods in Combinatorial Chemistry.’’ Technomic, Lancaster, 2000.

3
D. G. Schmid, P. Grosche, H. Bandel, and G. Jung, Biotechnol. Bioeng. Comb. Chem. 71,
149 (2001).
*Abbreviations: CLND, chemiluminescence nitrogen detection; C log P, calculated partition
coefficient; ELSD, evaporative light scattering detection; ESI-MS, electrospray ionization
mass spectrometry; FWHM, full width at half maximum; i.d., inner diameter; LC, HPLC,
liquid chromatography, high-performance liquid chromatography; LC/MS, liquid chroma-
tography – mass spectrometry; LC/MS/MS, liquid chromatography – mass spectrometry –
mass spectrometry; LC/UV/MS, liquid chromatography mass spectrometry with a UV
detector; LIB, compound library; log P, water/octanol partition coefficient; MUX,
multiplexed; RSD, relative standard deviation; SFC, supercritical fluid chromatography;
TFA, trifluoroacetic acid; TIC, total ion current; TOF, time of flight; TOFMS, time of flight
mass spectrometry.
4
H. Lee, L. Li, and J. Kyranos, Proceedings of the 47th ASMS Conference on Mass
Spectrometry and Allied Topics, Dallas, Texas, June 13–17, 1999.
[1] high-throughput LC/UV/MS analysis of libraries 3
Copyright 2003, Elsevier Inc.
All rights reserved.
METHODS IN ENZYMOLOGY, VOL. 369 0076-6879/03 $35.00
been used to achieve desirable high speed taking advantage of the low vis-
cosity of CO
2
.
5
However, the serial LC/MS approach by its nature does not
match the speed of parallel synthesis. Parallel LC/MS is the method of
choice to increase throughput while maintaining the separation efficiency.
An eight-probe Gilson 215/889 autosampler was incorporated into a
quadruple mass spectrometer.

6
This arrangement enabled the injection of
eight samples (a column from a 96-well microtiter plate) simultaneously
for flow-injection analysis/MS (FIA-MS) analysis to achieve a throughput
of 8 samples/min. A novel multiplexed electrospray interface (MUX)
7
was developed in 1999 and became commercially available for parallel
high-throughput LC/UV/MS analysis. The eight-way MUX consists of
eight nebulization-assisted electrospray ionization sprayers, a desolvation
gas heater probe, and a rotating aperture. It can accommodate all eight
high-performance liquid chromatograph (HPLC) streams at a reduced flow
rate of <100 l/min per stream and conduct electrospray ionization for all
eight streams simultaneously. Ions are continuously formed at the tip of
each sprayer and the MUX interface allows sprayers to be sampled sequen-
tially using the rotating aperture driven by a programmable stepper motor.
At any given time, only ions from one stream are admitted into the ion
sampling cone, while ions from the other seven sprayers are shielded. Each
liquid stream is sampled for a preset time with mass spectra acquired in full
mass range into eight simultaneously open data files synchronized with the
spray being sampled. With a 0.1-s acquisition time per sprayer and 0.05-s
intersprayer delay time, the time-of-flight (TOF) mass analyzer can acquire
a discrete data file of electrospray ion current sampled from each stream
over the entire HPLC separation with a cycle time of 1.2 s. Therefore,
this eight-way MUX-LCT was like having eight individual electrospray
ionization (ESI)-MS systems working simultaneously.
The MUX interface enables the coupling of parallel liquid chromatog-
raphy to a single mass spectrometer. This technology has had a great
impact in high-throughput LC/MS analysis. In drug development, a four-
way MUX interface was used on a triple quadrupole mass spectrometer
to simultaneously validate LC/MS/MS methods for the quantitation of

loratadine and its metabolite in four different biological matrixes
8
and of
5
M. C. Ventura, W. P. Farrell, C. M. Aurigemma, and M. J. Greig, Anal. Chem. 71, 2410
(1999).
6
T. Wang, L. Zeng, T. Strader, L. Burton, and D. B. Kassel, Rapid Commun. Mass Spectrom.
12, 1123 (1998).
7
V. De Biasil, N. Haskins, A. Organ, R. Bateman, K. Giles, and S. Jarvis, Rapid Commun.
Mass Spectrom. 13, 1165 (1999).
8
M. K. Bayliss, D. Little, D. M. Mallett, and R. S. Plumb, Rapid Commun. Mass Spectrom.
14, 2039 (2000).
4 analytical techniques [1]
diazepam in rat liver microsomes for in vitro metabolic stability.
9
The four-
channel LC/MS/MS system was also reported for the quantification of a
drug in plasma on both the narrow-bore and capillary scales.
10
By incorpor-
ating divert valves into this system, aliquots of plasma could be directly
analyzed without sample preparation. The four-channel LC/MS/MS has re-
duced method validation time, increased sample throughput by 4-fold, and
afforded adequate sensitivity, precision, and negligible intersprayer cross-
talk.
8,9
In protein analysis, an eight-way MUX coupled with a TOFMS

analyzer has proved to be a powerful tool to monitor the protein purifica-
tion process by screening fractions from preparative ion-exchange chroma-
tography with a throughput of 50 protein-containing fractions in less than
an hour.
11
A high-pressure gradient parallel pumping system (JASCO PAR-1500)
has been developed to conduct high-throughput parallel liquid chromatog-
raphy.
12
It is a 10-pump system where two pumps are used to generate a
binary gradient and eight pumps to deliver the mixed solvent to eight LC
columns. Comparing this system to a conventional system with two pumps
or a binary pump for LC gradient and a simple splitter to divide the gradi-
ent to eight LC columns, this system can ensure uniform flow rates through
each LC column. This system has been used for peptides and combinatorial
sample,
12
protein analysis,
13
and bioanalysis.
9
We have optimized an eight-way MUX coupled to a TOFMS analyzer
to carry out eight-channel parallel LC/UV/MS analysis of combinatorial
libraries
14
in the past 2 years. This system has not only provided the
capacity needed for library analysis, but also enabled simultaneous evalu-
ation of experimental parameters to expedite the method development
process. In this chapter, we discuss the optimization of this system and
present a high-throughput protocol for combinatorial library analysis. We

also compare the eight-channel parallel LC/UV/MS system to a conven-
tional single channel LC/UV/MS system in terms of performance and
operation.
9
D. Morrison, A. E. Davis, and A. P. Watt, Anal. Chem. 74, 1896 (2002).
10
L. Yang, T. D. Mann, D. Little, N. Wu, R. P. Clement, and P. J. Rudewicz, Anal. Chem. 73,
1740 (2001).
11
B. Feng, A. Patel, P. M. Keller, and J. R. Slemmon, Rapid Commun. Mass Spectrom. 15,
821 (2001).
12
D. Tolson, A. Organ, and A. Shah, Rapid Commun. Mass Spectrom. 15, 1244 (2001).
13
B. Feng, M. S. McQueney, T. M. Mezzasalma, and J. R. Slemmon, Anal. Chem. 73,
5691 (2001).
14
J. Zhao, D. Liu, J. Wheatley, L. Fang, and B. Yan, Proceedings of the 49th ASMS
Conference on Mass Spectrometry and Allied Topics, Chicago, IL, May, 27–31, 2001.
[1] high-throughput LC/UV/MS analysis of libraries 5
System Optimization
The high-throughput parallel LC/UV/MS system consists of an auto-
sampler with eight injection probes, two pumps for generating binary gra-
dient, eight UV detectors, and an eight-way MUX with a TOF mass
spectrometer. This two-pump arrangement keeps the system simple and
cost efficient. However, it does not provide pressure regulation for each LC
channel. To ensure flow consistency across each channel, we paid special
attention to the selection of tubing, joints, and columns. Columns are from
the same manufacturer and the same batch. The tubing is the same length
initially for each channel and is further adjusted by checking the flow at the

end. With these precautions, the flow from this two-pump system could be
split evenly among the eight channels. In addition, a standard mixture is
analyzed every 24 injections, and the retention times of these standards
are closely monitored to ensure an even flow across the eight channels.
Standards and Flow Monitoring
Five commercial compounds are chosen as standards in system opti-
mization and quality control (Fig. 1). Theophylline (log P 0.05), 5-phenyl-
1H-tetrazole (log P 2.41), reserpine (C log P 3.32), Fmoc-Asp(OtBu)-OH
(log P 4.43), and dioctyl phthalate (C log P 8.39)
15
were selected for our
15
L. Tang, W. Fitch, M. Alexander, and J. Dolan, Anal. Chem. 72, 5211 (2000).
N
N
N
H
N
O
O
H
3
C
CH
3
N
N N
NH
D Fmoc-Asp(OtBu)-OH
O

O
CH
3
CH
3
O
O
H
3
C
H
3
C
N
H
N
H
H
H
H
3
COOC
H
H
OCH
3
H
O
O
OCH

3
OCH
3
OCH
3
H
3
CO
O
N
H
O
O
HO
O
O
B 5-phenyl-1h-tetrazoleA Theophylline
E Dioctyl phthalate
C Reserpine
C log P = −0.39
log P = 0.05
C log P = 3.85
log P = 4.43
C log P = 8.39
C log P = 0.18
log P = 2.41
C log P = 3.32
Fig. 1. The structures of the five commercial compounds (A to E) used to monitor
performance.
6 analytical techniques [1]

standard mixture. Although only three compounds have experimentally
determined log P values (see above) their C log P values range from
À0.4 to 8.4, which covers most of the elution range for combinatorial
library compounds.
Without backpressure regulation for each channel, it is necessary to
minimize the flow rate fluctuation over time. The relative standard devi-
ation (RSD%) in retention time variation among the eight channels over
1 month for compounds A and B was less than 2% and for C and D it
was less than 1%. The RSD% for all channels over a 1-month period for
compounds A to D was 3.2, 2.4, 1.6, and 1.5%, respectively. Therefore, this
system is well suited for combinatorial library analysis. The UV chromato-
grams from channel 5 from different days are shown as an example in
Fig. 2A. The retention times of the four compounds (compounds A to D)
from all eight channels during a 1-month period are shown in Fig. 2B.
The throughput of this eight parallel LC/UV/MS system is 3200 com-
pounds per day for a 3.5-min cycle time per injection of eight samples
under current optimized conditions. It could be further increased by in-
creasing the gradient slope and flow rate. We have also determined that
five compounds in the standard mixture gave a linear response from 0.01
to 0.4 mg/ml.
16
16
L. Fang, M. Wang, M. Pennacchio, and J. Pan, J. Comb. Chem. 2, 254 (2000).
0.0
0.5
1.0
1.5
2.0
2.5
3.0

0 5 10 15 20 25 30
Time (day)
Retention time (min)
1.00
Retention time (min)
0.5 1.0 1.5 2.0 2.5 3.0
1.45
1.95
2.40
1 d
26 d
31 d
7 d
12 d
17 d
22 d
0.99
1.44
1.95
2.40
1.03
1.49
1.99
2.45
1.01
1.46
1.96
2.42
1.03
1.48

2.00
2.44
1.01
1.46
1.97
2.42
1.00
1.45
1.97
2.42
A
B
C
D
A B
Fig. 2. A selection of UV
214
chromatograms from channel 5 on different days and the
retention times from the eight-channel system on every day for standard compounds (A to D)
monitored over a 1-month period.
[1] high-throughput LC/UV/MS analysis of libraries 7
The T-Joint Position
A zero dead volume T-joint is used after each UV detector to split the
LC eluent to the MS analyzer and the waste to ensure a flow of 100 l/min
entering each channel. The position of the T-joint affected the separation
in the total ion chromatogram (TIC). When the T-joint was placed close
to the UV cell (320 mm from the UV cell), the distance between the
T-joint and the eight-way MUX interface is 780 mm. A sample had to
travel 12 s to reach the eight-way MUX inlet after the UV detector at a
flow rate of 50 l/min. Fmoc-Asp(OtBu)-OH had a full width at half max-

imum (FWHM) of 0.05 min when detected at the UV detector (Fig. 3C),
but the peak width was doubled at the position of the MUX inlet
(Fig. 3D). Such a peak broadening in TIC could have jeopardized product
identification. To minimize these effects, the T-joint was moved as close as
possible to the eight-way MUX inlet. With this modification the sample
reached the MUX inlet 2 s after leaving the UV detector. The FWHM in
the UV and TIC chromatograms were both 0.05 min (Fig. 3A and B). Thus
peak delay and broadening were all eliminated.
LC Conditions
Unlike a single LC/UV/MS system, reducing solvent consumption is im-
portant in this eight-channel LC/UV/MS system. A flow rate of 24 ml/min
on eight 4.6 Â 50-mm columns was used initially. This operation resulted in
a solvent consumption of 34.5 liters/day. To maintain the same separation
A
Retention time (min)
2.2 2.4 2.6 2.8
B
C
D
UV
214
UV
214
TIC
TIC
2.31
2.31
2.44
2.52
Configuration B

Configuration B
Configuration A
Configuration A
Fig.3.UV
214
and TIC chromatograms of compound D obtained from configurations A
(A and B) and B (C and D).
8 analytical techniques [1]
efficiency and minimize solvent consumption, columns with a smaller inner
diameter (i.d.), such as 2.1 mm, were evaluated. The standard mixture was
analyzed at flow rates of 6, 8, 10, 12, 14, and 16 ml/min. The LC gradient
was 10–100% B in 3.0 min and 100% B for 0.5 min. The chromatograms
in Fig. 4 show the separation results obtained from the original 4.6–mm-i.d.
column at 24 ml/min (A) and 2.1-mm-i.d. column at 8 ml/min (B), 12 ml/
min (C), and 16 ml/min (D). Below 8 ml/min, the very lipophilic dioctyl
phthalate did not elute. The separation was good at 12 ml/min and even
better at 16 ml/min. A flow rate of 12 ml/min with the 2.1-mm-i.d. column
was sufficient to maintain the separation efficiency, and consumed only half
the solvent.
For a particular library, an optimal LC column needs to be selected.
This parallel LC/UV/MS system could evaluate eight different columns in
4 min. Five C
18
columns of 2 Â 50 mm packed with 5-m particles made by
different manufacturers were evaluated simultaneously based on the separ-
ation efficiency of the standard mixture using trifluoroacetic acid (TFA) or
acetic acid as modifier. Chromatograms at 12 ml/min using 0.05% TFA
from each column are shown in Fig. 5. The Aqua column gave poor peak
shape for the early-eluting compound A. The Aqua and Luna columns
did not separate compounds B and C (data not shown) using 0.1% acetic

acid. The remaining three columns separated the five compounds well.
Retention time (min)
0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5
A
B
C
D
0.52
1.07
1.72
2.21
3.40
0.36
0.92
1.70
2.14
3.34
0.28
0.72
1.48
1.91
3.08
0.24
0.60
1.35
1.74
2.91
Fig. 4. UV chromatograms of the standard mixture at one of the eight channels separated
using a 4.6-mm-i.d. column at 24 ml/min (A), and a 2.1-mm-i.d. column, at 8 (B), 12 (C), and
16 ml/min (D).

[1] high-throughput LC/UV/MS analysis of libraries 9
An Efficient Rerun Protocol
In a high-throughput analysis mode, variable sample concentration
sometimes leads to inadequate or saturated signals and at times blocking
of the injector ports. Therefore, reanalysis of selected samples became a
necessity. However, due to the rigid design of the Gilson 215 Multiprobe
liquid handler, MUX-LCT cannot handle reanalysis efficiently. For
example, when an injector is blocked during an overnight queue, there
may be 24 failed samples on two 96-well plates (Fig. 6). Since the 8-probe
injector has to inject an entire column of eight samples for each run, it will
take 24 runs or 108 min to complete the analysis.
We have developed a new process to improve the efficiency of sample
reanalysis. This process includes four steps: data review, replating, reanaly-
sis, and data alignment. We have also generated an Excel template, a
Gilson’s Unipoint protocol, and an in-house visual basic program to
automate the process.
Raw data processed by the OpenLynx program was first reviewed for
consistency. Since the full scale of our analog channel is 2.1 Â 10
6
, samples
with a UV peak height over that limit saturate the UV detector. Besides
the detector saturation, samples can also overload the HPLC column and
generate broad peaks (FWHM > 0.1 min) in HPLC chromatogram. Both
types of signal saturation were identified, and a dilution factor was esti-
mated for each sample. Low sample concentration was another reason
for rerun. Failed external standards and sample carryover are indications
of an injector blockage. Samples with hydrophilic diversity eluted with
the solvent front using the generic method. These samples needed to be
A
Retention time (min)

0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5
B
C
D
E
Fig. 5. UV chromatograms of the standard mixture separated using 2.1-mm-i.d. columns
at 12 ml/min: Polaris (A), Zorbax (B), Omnisphere (C), Luna (D), and Aqua (E).
10 analytical techniques [1]
dissolved in water instead of methanol and should be analyzed using a shal-
lower gradient. All the above samples were entered into an Excel template
with a plate view. There are three output lists generated automatically by
the template: output to liquid handler for plating, output to MassLynx for
reanalyzing, and output for realigning the final data.
A Gilson liquid handler was programmed with the Unipoint software to
dilute and reformat failed samples. A volume of 120 l of solvent (metha-
nol unless specified otherwise) was added to each failed vial. The solution
was taken up and released by the sampling needle three times to ensure ef-
ficient mixing. A fraction of the liquid was then transferred to the target
plate. The fraction volume was determined on the Excel template by the
dilution factor, for example 40 l for 2:1 dilution. Solvent was allowed to
evaporate at ambient temperature from the new plate, and 200 l of solvent
was then added to each well in the plate. The new plate, reformatted and
compressed, was analyzed with the MUX-LCT system using the sample list
from the template. Using this new format, the 24 samples in the example
above (Fig. 6) were reanalyzed in three runs (13.5 min) instead of 24 runs
(108 min). This represents an 8-fold improvement in efficiency.
We have also created a visual basic program to modify sample location
information. Since sample location was hard-coded in the data file, samples
on the reformatted plate were of a different location from their original.
Once processed by OpenLynx, these samples will cause a ‘‘multiple injec-

tion conflict’’ with the samples that were originally in these locations. The
visual basic program used Microsoft scripting runtime objects to locate
each sample on the reformatted plate. It opened the header file, searched
for the sample location, and replaced it with the original location. This pro-
gram can also append customized information, such as a new identity after
‘‘cherry picking,’’ into the sample file. All added operations, such as plating
and alignment, were performed offline of the MUX-LCT. With this new
process, sample reanalysis became much more efficient.
123456789101112
123456789101112
123456789101112 123456789101112
A A
B B
C C
D D
E
F F
G G
H
H
Reformat
Alignment
Plate BPlate A Target plate
E
A
B
C
D
F
G

H
E
1917
21018
31119
41220
51321
61422
71523
81624
13 14 15 6 17 18 19 20 21 22 23 241
Fig. 6. Reformatting and realigning processes in reanalysis.
[1] high-throughput LC/UV/MS analysis of libraries 11
Combinatorial Library Analysis
In LC/MS analysis of combinatorial libraries, the MS determines the
product identity and its purity is determined by other on-line detection tech-
niques such as UV, evaporative light scattering detection (ELSD), and
chemiluminescent nitrogen detection (CLND).
17–20
UV detection is used
here to assess product purity based on the assumption of similar absorption
coefficients at 214 nm for the desired product and the side-products.
To develop a method for combinatorial library analysis, we first ana-
lyzed six to eight representative compounds from each library under gen-
eric LC/UV/MS conditions. These conditions would be used for library
analysis unless adjustments had to be made based on the study of these
representative compounds.
Evaluation of Representative Library Compounds
Five to eight representative compounds were evaluated simultaneously
due to the parallel nature of the system. Depending on the structure of a

library, this analysis was performed using acetic acid or TFA as modifier.
We found that the general LC gradient worked well for most of the library
except in a few cases in which very polar compounds eluted early. In these
cases, the sample solvent, solvent gradient, or LC column was varied to op-
timize the retention time. However, we had to adjust ion optics settings for
most libraries to ensure that the MH
þ
ion was the predominant ion to make
product identification simple. We found that sample cone voltage was a
critical parameter when all other ion optics parameters were kept constant.
This was reasonable because the sample cone separates the ionization
chamber with a pressure near atmospheric pressure from the vacuum
region with a pressure of a few Torr. Ions could be fragmented due to col-
lision with the gas molecules in this region. A higher sample cone voltage
would produce more energetic ions to undergo collision-induced dissoci-
ation. This eight parallel LC/MS system has dramatically accelerated this
process because up to eight compounds can be evaluated simultaneously
under the same experimental conditions.
Six compounds from library 1 (LIB1) have been analyzed simultan-
eously at sample cone voltages of 10, 20, 30, and 40 V. The mass spectra
of two compounds (LIB1-1 and LIB1-2) are shown in Fig. 7. Only MH
þ
17
L. Fang, M. Demee, T. Sierra, J. Zhao, D. Tokushige, and B. Yan, Rapid Commun. Mass
Spectrom. 16, 1440 (2002).
18
L. Fang, J. Pan, and B. Yan, Biotechnol. Bioeng. Comb. Chem. 71, 162 (2001).
19
D. A. Yurek, D. L. Branch, and M. Kuo, J. Comb. Chem. 4, 138 (2002).
20

E. W. Taylor, M. G. Qian, and G. D. Dollinger, Anal. Chem. 70, 3339 (1998).
12 analytical techniques [1]
100 200 300
400
500 600
700
800 900 1000
m/z
0
100
%
0
100
%
0
100
%
0
100
%
PFF115-299-45-A3-40V 86 (1.717) Cm (85:87)
TOF MS ES+
368
287.1
265.1
386.2
313.1
314.1
771.4
387.2

772.5
773.5
PFF115-299-45-A3-30V 87 (1.737) Cm (86:88) TOF MS ES+
438
771.4
386.2
265.1
261.2
287.1
387.2
772.5
773.5
PFF115-299-45-A3-20V 84 (1.677) Cm (83:85) TOF MS ES+
464
386.2
265.2
771.4
387.2
772.4
773.5
PFF115-299-45-A3-10V 91 (1.817) Cm (91:93) TOF MS ES+
416
386.2
771.5
387.2
772.5
773.5
100
200
300

400
500
600 700
800 900 1000
m/z
0
100
%
0
100
%
0
100
%
0
100
%
PFF115-299-40-3-40V 125 (2.487) Cm (124:126)
TOF MS ES+
499
423.2
316.1
290.1
317.2
424.2
845.4
PFF115-299-40-3-30V 125 (2.487) Cm (124:126) TOF MS ES+
480
423.2
316.1

424.2
845.4
PFF115-299-40-3-20V 122 (2.427) Cm (121:123) TOF MS ES+
383
423.2
845.4
424.2
PFF115-299-40-3-10V 127 (2.527) Cm (126:128) TOF MS ES+
165
423.2
845.4
424.2
846.4
20V
10V
30V
40V
20V
10V
30V
40V
A
B
Fig. 7. Mass spectra of LIB1-1 (A) and LIB1-2 (B) at sample cone voltage of 10, 20, 30, and 40 V.
[1] high-throughput LC/UV/MS analysis of libraries 13
[100% relative abundance (RA)] and 2MH
þ
(dimer, 50% RA) can be
found at 10 V for these compounds. Parent ions have been broken apart
as the sample cone voltage increases from 10 to 40 V. A major fragment

(m/z ¼ 316.1) with 70% RA could be detected in addition to MH
þ
(m/z
¼ 423.2, 100% RA) at 40 V for LIB1-1 (Fig. 7A). However, more extensive
fragmentation was observed for LIB1-2 (Fig. 7B). Four fragment ions
could be encountered along with MH
þ
(m/z ¼ 386.2, 90% RA) and
2MH
þ
(m/z ¼ 771.4, 85% RA) at 40 V. In terms of sensitivity, the total
ion counts for both of the compounds are lowest at 10 V and highest at
30 V for LIB1-1 and at 20 V for LIB1-2. In general, the higher cone voltage
produces the stronger ion intensity. However, higher cone voltage also
causes fragmentation, which in turn leads to uncertainty in product identi-
fication. As a compromise for six compounds, the sample cone voltage was
set to 20 V. The LC/MS chromatogram and mass spectra of all five
compounds under optimized conditions are shown in Fig. 8.
Six representative compounds from library 2 (LIB2) have also been ana-
lyzed to optimize the sample cone voltage. Mass spectra of two compounds
(LIB2-1 and LIB2-2) at sample cone voltages of 20, 30, and 40 V are shown
in Fig. 9.MH
þ
ions are shown as the predominant ions only at 40 V. Frag-
ment ions (m/z ¼ 378.3) could be observed with an RA of 100% and 80%
for LIB2-1 and LIB2-2 at 30 V. MH
þ
with 30% RA could be found as a
minor ion at 20 V while doubly charged ions with 100% RA were the major
ion. With a resolution around 5000, TOFMS made it easy to assign charge

states to each ion in the spectrum. Three ions with m/z of 234.6, 378.3, and
468.3 found from LIB2-1 at 30 V are displayed in the 3 amu window in Fig.
10A, B, and C, respectively. Charge states could be easily assigned based on
the mass difference between C12 and C13 for each ion observed in the mass
spectrum. A mass difference of a half unit indicated that the ion with m/z of
234.6 (Fig. 10A) has a charge state of 2 while ions of 378.3 and 468.3
have a charge state of 1 since a mass difference of one unit was observed.
It is concluded that product from LIB2 could be easily identified by a
doubly charged ion using a sample cone voltage of 20 V or identified by
a singly charged ion at 40 V. Detection sensitivity is higher for the doubly
charged ion at 20 V than that of the singly charged ion at 40 V. After
method development, a set of optimized ion optics settings was saved and
used for future analysis of the library along with the suitable LC conditions.
Library Analysis
Libraries were analyzed in 10 96-well plate batches. Each QC plate con-
tained 88 sample compounds. The last column of each plate was reserved
for sampling blank and standard controls. Standards were analyzed in
14 analytical techniques [1]
100 200 300 400 500 600 700 800 900 1000
m/z
0
100
%
0
100
%
0
100
%
0

100
%
PFF115-63-2-D4-MUX2 67 (1.595) Cm (67:69)
1: TOF MS ES+
483
457.1
399.1
913.2
458.1
855. 2
914.2
PFF115-63-1-A5-MUX2 58 (1.387) Cm (57:59) 1: TOF MS ES+
988
349.1
697.2
350.1
698.2
PFF115-45-C3-MUX2 46 (1.094) Cm (46:48) 1: TOF MS ES+
1.11e3
362.1
363.1
723.3
PFF115-45-A3-MUX2 70 (1.673) Cm (69:71) 1: TOF MS ES+
1.17e3
386.1
265.1
771.3
387.2
427.2
772.3

PFF115-40-3-MUX2 74 (1.761) Cm (73:75) 1: TOF MS ES+
917
423.1
845.3
424.2
846.3
0.50 1.00 1.50 2.00 2.50 3.00
Time
0
100
%
0
100
%
0
100
%
PFF115-63-2-D4-MUX2 1: TOF MS ES+
TIC
5.55e3
1.60
1.04
PFF115-63-1-A5-MUX2 1: TOF MS ES+
TIC
7.41e3
1.39
PFF115-45-C3-MUX2 1: TOF MS ES+
TIC
6.06e3
1.09

PFF115-45-A3-MUX2 1: TOF MS ES+
TIC
1.11e4
1.67
1.12
1.50
PFF115-40-3-MUX2 1: TOF MS ES+
TIC
7.41e3
1.76
0
100
%
0
100
%
0
100
%
Fig.8.UV
214
chromatogram and mass spectra of LIB1-1 to LIB1-5 under optimized conditions.
[1] high-throughput LC/UV/MS analysis of libraries 15
100 200 300 400 500 600 700 800 900 1000
m/z
0
100
%
PFF107-3-B4-a-1 48 (1.142) Cm (47:49)
1: TOF MS ES+

209
468.3
333.2
232.1
209.2
290.2
378.3
469.3
470.3
PFF107-3-B4-a 45 (1.070) Cm(44:47) 1: TOF MS ES+
968
378.3
234.7
235.2
468.4
379.3
469.4
PFF107-3-B4-a 45 (1.070) Cm(44:47) 1: TOF MS ES+
1.28e3
234.6
235.2
468.3
235.7
40V
30V
20V
100 200 300 400 500 600 700 800 900 1000
m/z
0
100

%
0
100
%
0
100
%
PFF107-3-D4-a-1 44(1.056) Cm (43:45) 1: TOF MS ES+
450
484.3
333.2
232.1
378.3
485.3
PFF107-3-D4-a 41 (0.984) Cm(40:42) 1: TOF MS ES+
647
484.4
378.3
242.7
243.2
394.3
485.4
PFF107-3-D4-a 40 (0.960) Cm(40:42) 1: TOF MS ES+
1.17e3
242.6
243.2
484.3
243.7
485.3
40V

30V
20V
A
B
0
100
%
0
100
%
Fig. 9. Mass spectra of LIB2-1 (A) and LIB2-2 (B) at sample cone voltage of 20, 30, and 40 V.
16 analytical techniques [1]
234 235 236 237
m/z
0
100
%
PFF107-3-B4-a 45 (1.070) Cm (44:48) 1: TOF MS ES+
370
234.7
235.2
235.7
378 379 380 381
m/z
0
100
%
PFF107-3-B4-a 45 (1.070) Cm (44:48) 1: TOF MS ES+
1.09e3
378.3

379.3
468 469 470 471
m/z
0
100
%
PFF107-3-B4-a 45 (1.070) Cm (44:48) 1: TOF MS ES+
459
468.4
469.4
A
B
C
Fig. 10. Isotope of three ions found from LIB2-1 at a sample cone voltage of 30 V. (A) Charge state of 2; (B, C) charge state of 1.
[1] high-throughput LC/UV/MS analysis of libraries 17
every 24 injections during analysis to monitor the performance consistency
of all eight channels. The analysis queue was constructed from an Excel
spreadsheet and imported into the MassLynx software for execution. After
acquisition, the data were processed using MassLynx in batches. Processed
data could be reviewed in OpenLynx by selecting a plate and clicking on
the desired well. The UV chromatogram and mass spectrum of the desired
product in LIB2, plate26, well D1, are shown as an example in Fig. 9.We
generated an Excel report that included filename, expected molecular
weight, purity of desired products at 214 nm, and a plate view with purity
indicated for all compounds in the 10-plate batch. Library LIB2 was
composed of 60 plates; it was analyzed in positive ion mode and processed
in six batches. The purity distribution of library LIB2 is shown in Fig. 11
with an average of 80.6% for 5280 compounds measured at 214 nm.
Figure 11 shows the plate view of all 60 plates. According to this protocol,
we have completed more than a half million LC/UV/MS analyses in a

period of 15 months with two eight-channel MUX-LCT systems.
Comparison of the Eight-Channel LC/UV/MS (MUX-LCT) System
with a Conventional Single-Channel LC/UV/MS System
The significant advantage of the parallel LC/MS system is its through-
put. Because eight LC/UV/MS analyses can be conducted simultaneously,
the total analysis time is decreased by a factor of eight. To analyze every
compound in a library of 2500 compounds at 3.5 min cycle time requires
146 h using a single channel LC/UV/MS system. However, it requires only
Average purity 80.6%
0
500
1000
1500
2000
10 20 30 40 50 60 70 80 90 100
Purity by UV
214
(%)
Number of compounds
Fig. 11. Library LIB2 purity distribution of 5280 compounds measured at UV
214
.
18 analytical techniques [1]
18.2 h to complete this task using an eight-channel parallel LC/UV/MS
system, and this makes it possible to perform LC/UV/MS analysis on every
compound for all of our libraries. In addition, this system also speeds
up method development because it simultaneously evaluates up to eight
parameters or variables such as the performance of eight different columns.
UV and TIC Chromatograms
An important concern in using an eight-way MUX interface is that the

acquisition cycle time (the time required to acquire one data point for each
channel) is longer, and the data acquisition time per channel is shorter,
than for a single-spray system. Therefore, the sensitivity might be lower
and the peak shape could be distorted. In our current system with a time-
of-flight mass spectrometer, the minimum time required for each acquisi-
tion cycle is 1.2 s with 0.1 s for data acquisition and 0.05 s for intersprayer
delay. The chromatographic baseline peak width was between 5 and 6 s in
the UV chromatograms and between 6 and 7 s in the TICs under general
LC/UV/MS conditions. A maximum of five MS data points could be ac-
quired to define a peak, which resulted in slightly distorted peak shapes
in the TICs. On the other hand, peak shapes were much better defined in
a single-channel system because more than 10 data points could be easily
obtained. For combinatorial library analysis, lower sensitivity is not a prob-
lem because the parallel synthesis method always produces enough com-
pound for analysis. The limited number of data points across an LC peak
was usually not a problem because the MS data were used only to identify
the peak of interest. In theory, one or two data points (TOF mass spectra)
should be sufficient to confirm the expected molecular weight. The product
purity was obtained from the UV chromatogram, where the number of
data points was sufficient to ensure excellent peak shape and precision.
Data Acquisition Using Positive and Negative Ionization
In a single-spray system, it is common to analyze samples in both posi-
tive and negative ion modes by switching polarity during a single data ac-
quisition. This practice makes the best use of precious MS time and
identifies products by their presence in both positive and negative ion
forms. Both positive and negative ESI modes are available for the eight-
channel MUX-LCT system. However, the polarity change within a single
data acquisition would make the cycle time much longer. Therefore, we
prefer to analyze samples using a single polarity, and conduct a separate
experiment with the other polarity if necessary. With this arrangement,

high-throughput LC/MS analysis with both positive and negative mode
is available.
[1] high-throughput LC/UV/MS analysis of libraries 19
Sample Rerun
For a conventional single-channel LC/UV/MS system, a single un-
satisfied well could be easily reanalyzed. In the eight parallel LC/UV/MS
system, the rerun procedure was different from that of the single-spray
system. If problems were found in a single channel, such as retention time
shift or channel blockage, 12 wells in a row would fail and the whole plate
had to be reanalyzed. We have developed a rerun protocol that made the
parallel LC/MS analysis as efficient as the single-channel system.
Operation and Maintenance
In the eight-channel parallel LC/UV/MS system, a standard mixture
was analyzed every 24 injections. This was indispensable for the operation.
The variation of the retention time across eight channels was monitored
closely to ensure consistency for the eight channels. A significant retention
time shift indicated problems that usually could be overcome by replacing
the frit in the precolumn filter. A diminished peak area or a change in peak
shape of standards indicated column deterioration. We started with eight
columns from the same batch for sample analysis. Deteriorated columns
were replaced individually. This practice gave us satisfactory analysis data
for combinatorial library analysis with minimal cost.
We anticipated difficulty in maintaining and troubleshooting an eight-
channel parallel system because the problems in the autosampler, LC
columns, UV detectors, and MS interface would be multiplied by eight.
In fact, with the convenience of simultaneous analysis of the other seven
channels, the diagnosis and troubleshooting were made easier. The com-
plete system was easily divided into four functions: injection, separation,
UV detection, and MS detection. By running the standard mixture on eight
channels then switching channels at different function sites and rerunning

the standard mixture, problems were easily isolated. Fixing the problems
was exactly the same as for the single-spray system.
Conclusion
We have optimized an eight-channel parallel LC/UV/MS (MUX-LCT)
system for high-throughput LC/UV/MS analysis of large combinatorial lib-
raries. Since the LC gradient is divided into eight LC columns by a simple
splitter, the flow fluctuation has been continuously monitored and minim-
ized using a standard mixture during analysis to ensure performance con-
sistency among the eight channels. To preserve the separation integrity in
the total ion chromatogram, the zero dead volume T-joint used to split the
flow (after UV detection) should be best placed as close to the eight-way
20 analytical techniques [1]