Report
Affinity NMR A new drug-screening tool that probes ligandreceptor interactions. In the traditional drug discovery method, candidates are synthesized, purified, and tested individually—a process that takes time. Time can be translated into money, and that has become an issue of major concern in pharmaceutical industries. As a result, combinatorial chemistry, which is expected to generate a vast array of diverse chemicals rapidly and efficiently, has swept through the drug industry like a brushfire through the dry chaparral (1,2). One of the challenges of the combinatorial approach is the screening of large quantities of compounds to identify drug leads with a desired biological activity or with a binding affinity for a specific target Thus highthroughput biological screening has become indispensable in searching for "hits" combinatorial libraries However beused in screening nonspecific binding of several mixture components in an assay can cause proulems in man bi l cr' al s stems For e
ample, the observed false activity may be the sum of several weak interactions. Accordmgly, effective analytical approaches are needed to characterize interactions as 11
i
• 1
J..JL
il
well as to identify the components responsible for the desired activity. NMR is an excellent tool for analyzing mixtures. Monitoring chemical shift perturbations of protein 1SN-1H heteronuclear single-quantum correlation (HSQC) data—a two-dimensional map of 15N and 1 H chemical shifts correlated by their couplings has already been widely used •11
e
,1
i
•
i
Aidi Chen Michael J . Shapiro Novartis Institute for Biomedical Research Analytical Chemistry News & Features, October 1, 1999 669 A
Report to characterize interactions between proteins and various ligands (3). However, incorporating NMR into the front line of drug screening is largely impeded by its slow turnaround time for providing structural details. Enormous efforts have been made in adjusting NMR techniques to address this time issue. Methods that can directly detect and identify active components from mixtures could eliminate the time-consuming step of deconvoluting the mixture by performing a series of individual experiments and reduce the "false positives" resulting from an additive effect of individual small activities This Report describes affinity NMR also known as diffusion-edited NMR which identifies compounds that bind directly to a lartre receptor mise aid for intprpretinff hieh-throuffhrmt arrppninp results (4)
we need to examine the diffusion experiment. Each spin in a magnetic field precesses along the field at a speed determined by the magnetic field strength. If a magnetic field gradient whose field strength is position-dependent is superimposed on the main magnetic field, then the precessing frequency of each spin will also be position-dependent. Such a spatial labeling scheme is what diffusion NMR relies on to detect the translational movement of molecules as illustrated in Figure 1. Each molecule is first labeled according to its spatial position by a magnetic pulse field gradient during the gradientencode period (T) . Then a diffusion time window (A) is opened to record the random movement of molecules. Before the NMR signal is detected, a gradientdecode period (?) is employed to decrypt the labeling code so that the information
Figure 1. Block diagram of the diffusion experiment.
What is affinity NMR?
As implied by its name, affinity NMR is meant to distinguish active ligands that show binding affinities to a particular receptor in a pool of nonactive ones. With no need for physical separation of the mixture or additional deconvolution steps, affinity NMR can increase the efficiency and accuracy of drug screening. The basis for the "separation" in this approach is that the diffusion rate of a small ligand will change considerably when it complexes with a large molecule. As a result, the diffusion coefficients of binding and nonbinding ligands can differ substantially, which allows the diffusion NMR technique to be used to selectively detect binding ligand signals. This concept of separating compounds by receptor affinity is reminiscent of separation by affinity chromatography. To understand how affinity NMR works, 670 A
about molecular movement can be interpreted through signal attenuation. (Strictly speaking, the gradient encode and decode periods have some overlap with the diffusion time window.) In the diffusion experiment, series one-dimensional (1-D) NMR spectra are acquired with different gradient values. Each NMR signal (I) decays exponentially as the gradient (K) increases. 7 is a function of the diffusion coefficient (D), the gradient used in encoding and decoding the molecule, and A such that (5) z
I{A) = 7(A = 0)exp[-X" D(A - -)]
(1)
in which K = ygb, y is the gyromagnetic ratio, g is the gradient strength, and 8 is the duration over which gradient is applied. By plotting I versus A2(A - 8/3) and fitting the exponential decay curve, the diffusion coefficient of each molecule
Analytical Chemistry News & Features, October 1, 1999
in the system can be determined. Because of its convenient and nondestructive nature, pulsed-field gradient diffusion NMR has become a popular tool for applications such as resolving complex mixtures (6), probing molecular interactions (7), and measuring association constants (8). Even molecular weight distribution of compounds can be estimated from the measured diffusion profile (9,10). The diffusion NMR technique has been applied to drug screening and is termed "affinity NMR". A complete diffusion measurement, which is carried out in a two-dimensional (2-D) fashion, provides accurate information about the system. However, this process can be time-consuming and, for the fast-paced industrial setting, offers detailed quantitative information that may not be essential. For drug screening, sometimes it may not be necessary to measure D to differentiate binding ligands from nonactive ones. As long as the binding ligands show sufficient difference in diffusion properties from the nonbinding compounds a 1-D affinity NMR spectrum may be sufficient to discriminate active ligands (slow-diffusing species) from nonactive components (fast-diffusing SDecies) by a diffusion filter with an appropriate gradient The first demonstration of the technique involved a receptor (hydroquinine 9-phenanthryl ether), which can be thought of as a bulky guy with eight small potential ligands (4). Who is willing to "dance" with him? In the normal proton spectrum, all nine compounds are present (Figure 2a). As a ligand complexes with the bulky receptor, its movement is slowed. When a diffusion filter is applied in the affinity NMR experiment tofilteraway the fastdiffusing components, only the two slowmoving receptor-binding ligands and the receptor are left in the spectrum. All the other ligands have disappeared. In this case, the two binding ligands can be easily identified by their characteristic chemical shifts (Figure 2b) Because binding is detected by observing the ligand NMR signals, affinity NMR can be "tuned" to a desired sensitivity level. It has been demonstrated that by changing the concentration of the recep-
Figure 2. (a) 1 -D 4 0 0 - M H z 1 H NMR spectrum versus (b) diffusion-edited spectrum for a nine-component mixture in CDCI 3 . The concentration of each component is 10 mM. The components are (1) DL-isocitric lactone, (2) (S)-(+)-0-acetylmandelic acid, DL-N-acetylhornocysteine thiolactone, (±)-sec-butyl acetate, propyl acetate, isopropyl butyrate, ethyl butyrylacetate, butyl levulinate, and hydroquinine 9-phenanthryl ether, (b) Diffusion-edited 1H NMR spectrum showing chemical shifts arising from compounds 1 and 2. All other shifts are from the 9-phenanthryl ether.
tor, the binding ligands show up sequentially in their order of binding affinity (11). In another case involving 10 peptides and the receptor vancomycin, the affinity NMR signal intensities from the three binding peptides were found to be concordant with their relative binding strengths (12). These experiments demonstrate the potential of affinity NMR to rank ligands by their binding affinities. The rank-order binding by affinity NMR is not without caveats. It is assumed that the stronger binders spend more Figure time with the receptor and therefore appear to diffuse more slowly than the weaker binders. (This tendency is true only if all the ligands have the same diffusion rates when they are free and if they diffuse at the same rate when they are complexed with the receptor.) However, a slower diffusion rate (hence a slower signal decay rate) for a molecule does not guarantee it will show a stronger signal than a faster dif-
binding affinity can be obtained by a complete diffusion measurement of the ligands and receptor (8,12). Using the diffusion NMR technique, we have seen that a small mixture of compounds can be selectively edited to detect and identify components involved in molecular interactions. Even the structure of the interacting ligand can be determined with powerful multidimensional NMR techniques (13). The system can also be "tuned" to the binding affinity of the ligand, or the ligands can be rank-ordered qualitatively by binding affinity according to the signal intensities. No physical separation and data deconvolution steps are neceswhich suggests that affinity NMR is an effective tool for binding analysis and molecular-recognition studies However, in reality, life is not that simple. A lot of factors can complicate the analysis. In the following sections we will discuss some problems we have encountered and some possible solutions. Interference of receptor signal
One potential problem with affinity NMR is that the receptor signals are always present in the diffusion-edited spectrum, and this condition can make it difficult to identify fusing species in the affinity NMR binding ligand signals. An isotope-filtered spectrum. affinity NMR (14) experiment was develFor example, methyl protons (CH3) oped to simplify the spectrum (Figure 3). of a weak-binding ligand may still have a The new experiment first employs the higher intensity than a methine proton diffusion-editing technique to eliminate (CH) of a strong binder in a diffusionsignals arisingfromthe nonbinding comedited spectrum. An analogy can be pound as before. Then a 13C isotope filter, drawn to a container leaking water. A whichfilterssignals from protons directly attached to 13C, is applied to suppress receptor signals. Obviously this technique requires a receptor that is fully 13C isotope labeled The experiment was demonstrated by using i3c/15N-labeled 3 , Block diagram of isotope-filtered affinity NMR. stromelysin catalytic domain. With isotope filtering, an affinity NMR slower draining system does not guaranspectrum was obtained containing only the tee that the container has more contents bound ligand signals. left than one with a larger draining hole. It depends on the initial contents. When different compound signals are being compared for order-ranking in the affinity NMR, their initial signal intensities (with no gradient applied) must be taken into account. A more accurate evaluation of
Another approach to suppressing receptor signals, which does not require isotope-labeled material, is to directly detect nonbinding compounds (15). This method relies on the spectral subtraction of two diffusion-edited NMR data sets,
Analytical Chemistry News & Features, October 1, 1999 671 A
Report
Figure 4. Chemical exchange effects on affinity NMR. Affinity NMR with LED pulse sequence (a, b) compared with affinity NMR with bipolar LED pulse sequence (c, d). (a) and (c) Precessing of magnetic vector in the xy plane during gradient-encode and -decode periods in the cases of no exchange and with exchange, dynamically portrayed by the black arrows. The green arrows and the red arrows (arrows overlap in [c]) indicate the vector position at the beginning and the end of each period, respectively. The red dash arrow in (a) illustrates the actual signal vector being detected. The orange dash arrows in (c) show the vector position before and after each 180° pulse, (b) and (d) Stack plot of 1H spectra of the vancomycin and DDFA mixture (molar ratio of 1:2.8) as T increases.
which were acquired at low gradient strength (full spectrum with every compound signal) and high gradient value (spectrum with binding ligand and receptor signals). The resulting difference spectrum contains only nonbinding components. A spectrum with binding ligand signals can be obtained by another spectrum subtraction from a spectrum detected in the absence of receptor. Chemical exchange
Chemical exchange modulation can critically affect the ability of affinity NMR to observe binding ligand signals (16). To understand the effect of chemical exchange modulation on the signal, we need 672 A
to take a closer look at the most-often-used experiments: stimulated echo (STE) (17) and longitudinal eddy-current delay (LED) (18) pulse sequences. In addition to molecular translational motion (diffusion), each nucleus rotates along the main magnetic field (e.g., 2-axis). During the gradient-encode and -decode periods with all the magnetization tilted into the xy plane (perpendicular to the main magnetic field), each nuclear spin will precess along the z direction at a frequency characterized by its chemical shift frequency. Different spins with varying chemical shifts will spread out at the end of the gradient-encode period. (We do not need to worry about spin rotation during the diffu-
Analytical Chemistry News & Features, October 1, 1999
sion time, as all the magnetization is storei along the 2-axis.) To avoid this situation, the phase angle (relative to the Ar-a#is) of each spin acquired during the gradient-encode delay is offset by precessing in the opposite direction during the gradient-decode period. The two delays—gradient-encode and -decode delay T—are set to be the same so that all of the spins can be refocused to their starting position (z-axis). With this arrangement of course the spins must be precessing at the same speed during the gradient-encode and -decode period (Figure 4a) However, if the spin takes off with the speed of a "train" (coA) during the gradi-
ent-encode pcnod, it will not return to its starting position by taking a "bike" (roB) during the gradient-decode delay. This situation may occur if there is a chemical exchange within the diffusion time window, and spin ends up on a different site (with a different chemical shift frequency) in the gradient-decode period than it had in the gradient-encode time. | ) to the Therefore, the relative position ( #-axis depends on the difference between the two speeds (coA-coB) and the delay (x) of gradient encoding and decoding. If the detector is placed to observe the signal along its starting position (e.g., #-axis), then the intensity of the signal detected (i.e., the projection of the vector on the jc-axis) will be modified by cos[(coA-(BB)x]. This signal intensity modulation phenomenon is exactly what is observed for a series of LED experiments with a mixture of vancomycin and the tetrapeptide AspAsp-Phe-Ala (DDFA), as the x increases (Figure 4b). DDFA interacts with vancomycin, exchanging between free and bound states. The striking effect is that the methyl protons of bound DDFA and the methyl group of free DDFA not only decay with respect to relaxation and diffusion as the x increases, but they also oscillate at the frequency. As we expected, the signals are cosine-modulated and the oscillation frequency is the chemical shift difference between the two methyl signals On the other hand the peaks from the vancomycin arefreefromthis kind of oscillation This signal oscillation can cause serious problems in affinity NMR. At certain x values, some signals can totally disappear. If these x's were chosen in the affinity NMR experiment, then some signals would never be observed independent of what gradient is applied. This tendency can be very destructive in affinity NMR, because the experiment depends on observing ligand signals to identify active components from a mixture. This problem will present itself in all diffusion experiments that involve LED or STE pulse sequences and will cause serious signal distortion and sensitivity loss Even negative signals are possible This problem can be solved by using the bipolar LED (19) (or bipolar STE) sequence. Besides the replacement of one
monopolar gradient with two gradients of opposite polarity, the important modification of the program arises from a 180° pulse inserted in the middle of both the gradient-encode and the gradient-decode period. The 180° pulse facilitates the refocusing of the magnetization within each period before it gets any chance to jump between different sites. Therefore, this experiment does not need the gradientdecode period to offset the phase angle acquired during gradient-encode delay.
Chemical exchange can really blur the boundary between ligands and nonbinding compounds. So even if the spin takes a train during the gradient-encode period and a bike during the gradient-decode period, it always returns to its starting position (#-axis) at the end of each duration (Figure 4c). Therefore, if the detector is placed to observe signal along *-axis, the maximum signal intensity is always detected. Now let's do the same experiment with a bipolar gradient pair (Figure 4d). The chemical shift modulation is completely removed. The signals from the bound and the free DDFA now decay with the time x in the same manner as the vancomycin peaks, although they are still involved in chemical exchange. Clearly the bipolar LED (or bipolar STE) sequence can totally negate this exchange effect and retain diffusion information without intensity distortion. Along with less eddy-current effect and better resolution (19) bipolar LED or bipolar STE offers a cleaner spectrum which isfreeof chemical exchange modulation and whose results are not prone to misinterpretation
NOE pumping
Besides exchange modulations, other complications arise in the affinity NMR experiment. Because binding ligands usually exchange between free and bound states, the observed diffusion coefficient (Z)obs) is the weighted average of the free and bound diffusion coefficients, Dfree and Dhoand, obs=
rreA
free
bound bound
@)
in which % ee and *bound are fractions of the free and bound states, respectively, and % ee + xhoani = 11 When n amall ligand binds to a macromolecule, Z)free can be as much as 1 or 2 orders of magnitude larger than Dbound. Clearly, the free ligand can be weighted much more heavily than the bound ligand, and the observed diffusion coefficient can be closer to the Dfree than to the Db d. Chemical exchange can really blur the boundary between binding ligands and nonbinding compounds. As a result the difference in diffusion coefficients between the nonbinding and the binding ligands could be too small to be distinguished Another difficulty arises from line broadening. It is possible that by changing the relative concentrations of the ligands and the receptor (hence, % ee and Abound)- the observed diffusion coefficient of the binding ligand is closer to DhounA. In this case, the difference between nonbinding and binding ligands in diffusion rate is sufficient, but the ligand signals left in the affinity NMR spectrum are dominated by the binding ligand in the bound state. When a ligand complexes with a big molecule, the bound ligand signals could become too broad to be observed directlv To demonstrate the difficult condition, we used human serum albumin (HSA) as a receptor, salicylic acid as a binding ligand, and glucose and ascorbic acid as the nonbinding compounds. The apparent diffusion coefficients of the three ligands are not very different. Although the diffusion coefficient of salicylic acid is slowed by complexation with HSA, the difference is not significant because of chemical exchange averaging. Applying a strong gradient to suppress the nonbinding compound signals in the affinity NMR experi-
Analytical Chemistry News & Features, October 1, 1999 6 7 3 A
Report
Figure S. (a) Block diagram of NOE pumping experiment, (b) Normal 1 H NMR spectrum of HSA with three potential ligands. (c) The evolution of 1-D 1H spectra of the mixture with mixing time tm in NOE pumping experiment. S is salicylic acid, A is ascorbic acid, and G is glucose.
ment caused the binding ligand signals to disappear as well. An important difference between the binding and nonbinding compounds is that the binding ligand can be intimately related to the receptor. Can this closeness be a basis for separation? To obtain distance information, we resorted to nuclear Overhauser effects (NOEs). NOE is caused by a dipole-dipole interaction among spins, leading to a transfer of magnetization between dipolar-coupled spins (20). NOE is important to NMR spectrometrists because it can provide unique information about through-space interactions that is not easily extracted by other techniques Hence the NOE experiment has become an indispensable tool for lecular structure determination To differentiate between binding and nonbinding ligands, a new experiment was proposed: NOE pumping (21). The experiment is shown in Figure 5a. A diffusion filter is first applied to prepare a state in which only protein signals can be detected. (In fact, any experiment that can suppress the ligand signals while preserving the macromolecule signals can be used. For example, if the receptor or the
ligands are isotope-labeled, an isotope filter can be effective.) Then, an NOE experiment is applied to allow the signal redistribution from protein to binding ligands. The pumping mechanism can be visualized as follows. At the beginning of the NOE experiment, all of the ligand signals are destroyed, while the protein signals are conserved. During the mixing time tm, the magnetization is being transferred. When a protein interacts with a binding ligand, magnetization transfer might occur if the interacting molecules stay close enough for enough time. The consequence of this intermolecular NOE is that the observable magnetization is transferred from the protein to the binding ligand and the protein signals become greatly attenuated Therefore only those ligands that are capable of binding to the protein can be detected As the ligand disassociates from the protein, the transferred polarization is carried away and conserved by the longer relaxation time of the ligand in the free state. Here, the macromolecule signals are serving as a big reservoir that delivers detectable magnetization only to the ligands that
bind to the receptor. It seems that the signal is continuously "pumped" from the protein to the binding ligands by intermolecular NOE. And, because magnetization can be transferred to the binding ligand in the free state, signals from the binding ligand (average of free and bound) can still be detected even if the bound ligand signal is too broad to be detected directly. The diffusion-assisted NOE pumping experiment is demonstrated in Figure 5c, which shows the evolution of 1-D NMR spectra with tm. At very short tm, the spectrum contains only protein signals. But as tm increases, the salicylic acid signals— including the macromolecule-associated water signal—grow substantially as the protein signals decrease. Apparently, the HSA signals are pumped to the molecules that bind to the protein. However, the nonbinding glucose and ascorbic acid remain undetectable (compared with Figure 5b). With the NOE pumping experiment the binding ligands be clearly distinguished from the nonbinding NOE pumping provides a complementary method to affinity NMR and offers several advantages over conventional NMR screening approaches used to identify macromolecular ligands. Because the binding ligand signal is transferred from the protein, this technique is quite resistant to false positives. Chemical shift, relaxation, and diffusion parameters used in conventional NMR screening approaches (22) are sensitive to sample environment changes as well as binding activities. Drawing the line between parameter changes by binding activities and by other chemical environment variations such as temperature or pH
be diffi-
cult Because the ligand signals are monitored this method can be applied to very larfre proteins without well-resolved chemical shifts and without isotnnp lahplfavors large receptors T h e sensitivity of the technique increases with t h e molecular weicrht of
the receptor, because the polarization transfer rates are directly proportional to the correlation time (or the size of the molecule) in the slow-motion limit. In addition, broader macromolecular signals Ml 1
1
1 - 1 1 . 1
1 - 1 -
1-
will be less likely to obscure binding ligi
•
i
^i
i
i
i
and signals. The approach can also be 674 A
Analytical Chemistry News & Features, October 1, 1999
applied to the study of weakly bound ligands such as solvent. Furthermore, detailed binding site information is possible by extending the NOE experiment into multidimensions. Another efficient NMR screening technique based on intramolecular transferred NOE (23) is worth mentioning. Although NOE pumping and transferred NOE screening techniques are both rooted in the NOE, the separation basis for transferred NOE is the tumbling rate rather than the proximity. A small, free ligand with a fast tumbling rate generally exhibits positive intramolecular NOEs. However, if the ligand binds to a large molecular mass receptor, then it tumbles with the receptor at a much slower rate and may show strong negative NOEs. When chemical exchange is present between free and bound states the strong "transferred" to the sharrj resonances of the free ligand and can be easily detected (24) Therefore intramolecular NOEs binding- ligand can be orjoosite to those of the small nonbindine molecules 2-D transferred NOE experiments, which reveal any sign changes in the intramolecular NOEs of compounds following addition of a receptor, offer another effective way to differentiate binding from nonbinding ligands. Unlike NOE pumping, the transferred NOE screening technique does not need intermolecular NOEs between the ligands and the receptor, and the technique requires less receptor. However, NOE pumping can be less time-consuming because it is a 1-D experiment and can be applied to larger ligands, which may give negative intramolecular NOEs even in the free state Summary
Because binding ligands are directly detected and identified, the diffusion-based NMR method and NOE pumping approach promise to greatly simplify deconvolution in drug screening. An additional advantage of these techniques is that low-affinity ligands, which might be missed by highthroughput screening, can be detected and could serve as synthetic precursors for higher affinity ligands. The biggest challenge to NMR meth-
odology lies in its sensitivity. Compared with other techniques, such as MS (25), NMR methods for screening mixtures are limited by their relative insensitivity. Because of issues such as solubility, stability, and mass limitation, it is not in general judicious to simply increase the concentration of the mixture. Improvements in hardware and software are necessary to extend the applicability of the affinity NMR method to the screening of larger and more complex mixtures. A boost in
(9) (10) (11) (12) (13) (14) (15) (16)
NMR will continue on its fast track of development in support of drug discovery.
(17) (18) (19) (20)
(21) (22) (23) (24)
sensitivity and screening capacity of NMR technique is possible by the implementation of microcoil (26) and flow probe techniques. An upsurge in the capabilities of mixture analysis could be achieved with a combination of independent and complementary techniques (e.g., HPLC, MS) (27). As a unique, nondestructive, and versatile tool, NMR will continue on its fast track of development in the support of drug discovery.
(25) (26) (27)
Kuchel, P. W. Biophys. J. 1994, 67, 20963009. Chen, A; Wu, D.; Johnson, C. S., Jr./. Am. Chem. Soc. 1195,117, 7965-70. Jerschow, A; Mueller, N. Macromolecules 1998,31, 6573-78. Lin, M.; Shapiro, M. J.; Wareing, J. R. /. Org. Chem. 1997, 62, 8930-31. Bleicher, K.; Lin, M.; Shapiro, M. J.; Wareing, J. R./. Org. Chem. 1998, 63, 8486-90. Lin, M.; Shapiro, M.J./. Org. Chem. 1996, 61, 7617-19. Gonnella, N.; Lin, M.; Shapiro, M. ..; Wareing, J. R.; Zhang, X./. Magn. Reson. 1998, 131, 336-38. Hajduk, P. J.; Olejniczak, E. T; Fesik, S. W. /. Am. Chem. Soc. 1197,119,12257-61. Chen, A; Johnson, C. S.. Jr.. Lin, M.; Shapiro, M.J./. Am. Chem. Soc. 1998,120, 9094-95. Tanner, J. E.J. Chem. Phys. 1970,52, 2523-26. Gibbs, S. J.; Johnson- C. S., Jr./ Magn. ReR son. 1991, 93, ,95-402. Wu, D.; Chen, A; Johnson, C. .., Jr. / Magn. Reson., Ser. A 1995,115, 260-64. Ernst, R. R.; Bodenhausen, G.; Wokaun, A Principles of Nuclear Magnetic Resonance in One and Two Dimensions; Oxford University Press: New York, 1987; Chapter 9. Chen, A; Shapiro, M. J./. Am. Chem. Soc. 1998,120, 10258-59. Otting, G. Curr. Opin. Struct. Biol. 1993, 3,760-68. Meyer, B.; Weimar, T; Peters, T Eur, .J Biochem. .997,246, ,05-099 Clore, G. M.; Gronenborn, A M.J. Magn. Reson. .982,48, 402217. Van Breemen, R. B.; Huang, C; Nikolic, D.. Woodbury, C. P; Zhao, Y.; Venton4 D. L. Anal. Chem. 1197, 69, 2159-646 Olson, D. L.; Lacey, M. E.; Sweedler, J. V. Anal. Chem. 1198, 70, 252 A-264 A. Shockcor, J. P.; Unger, S. E.; Wiison, I. D.; Foxall, P.J.D.; Nicholson, J. K.; Lindon, J. C.Anal. Chem. 1196, 68, 4431-35.
Aidi Chen received her Ph.D. at the University of Nortt Carolina ana is currently a postdoctoral fellow at the Novartis Institute for Biomedical Research. Her research interests focus so the development of tew NMR techniques snd their application on drug screening, and the characterization of References molecular interactions by NMR. Michael (1) Baum, R. Chem. Eng. News 1996, 74 Shapiro iishe NMR technology lealer at (Feb 12), 28. the Novartis snstitute for Biomedical Re(2) Czarnik, A. W. Anal. Chem. 1198, 70, search, where he has been named a Novar378A-386 A tis Leading Scientist. His research interests (3) Shuker, S. B.; Hajduk, P. J.; Meadowss include the use of NMR in combinatorial R. P.; Fesik, S. W. Science 1916,274, 1531-34. chemistry method development for chiral (4) Lin, M.; Shapiro, M. J.; Wareing, J. R. purity determination and new NMR meth/. Am. Chem. Soc. 1997,119, 5249-50. (5) Stejskal, E. 0.;Tanner, J. E.J. Chem. Phys. ods for studvins molecular interactions Address correspondence about this article 1965, 42, 288-92. to Shapiro Core Technologies Novartis (6) Morris, K. F.; Johnson, C. S., Jr./. Am. Chem. Soc. 1193,115, 4291-99. Institute for Biomedical Research Novartis (7) Stilbs, P. Prog. NMR Spectrosc. 1987,19, Pharmaceutical Corp Summit Nf 079011-45. (8) Lennon, A. J.. Scottt N. R.; Chapman, B. E.; 1398 Analytical Chemistry News & Features, October 1, 1999 6 7 5 A
