Strategies and Data Precision Requirements for the Mass

---

Anal. Chem. 1997, 69, 4354-4362

Strategies and Data Precision Requirements for the Mass Spectrometric Determination of Structures from Combinatorial Mixtures Karl F. Blom

Experimental Station, The DuPont Merck Pharmaceutical Company, P.O. Box 80500, Wilmington, Delaware 19880-0500

Mass spectrometric data can be obtained for compounds in bead-bound combinatorial mixtures by several techniques. However, little specific information is available regarding (1) how well these data differentiate between candidate structures in large combinatorial pools, (2) what precision of data is required to achieve adequate specificity in these analyses, and (3) what are the best strategies for applying these data. In this work, computer modeling is used to address these questions. Strategies employing multiple filters (i.e., those that differentiate possible structures using more that one measured mass spectral parameter) are found to provide better specificity and to be more robust (that is, the specificity is less dependent on the precision of the data) than discrete filters. With moderate precision data (e.g., 50 ppm mass precision, 10% isotope ratio precision), multiple filter strategies are found to give unequivocal results for ∼80% of the populations of combinatorial mixtures with most of the remaining degeneracy at the 2-fold level. A simple protocol for the application of multiple filter methods is presented. Combinatorial chemistry is a powerful tool for identifying lead compounds in drug discovery, and numerous approaches are being developed throughout the pharmaceutical industry. While there does not yet seem to be a clear consensus as to the best approach, one-bead, one-structure-based approaches1 are finding wide use. All of these strategies contain four common elements: (1) synthesize libraries of one-bead, one-compound mixtures by the split and pool method; (2) screen libraries to identify beads containing active compounds; (3) determine identity of compounds on active beads; (4) resynthesize and screen putative active compounds to confirm activity. A productive combinatorial discovery program may require the synthesis and screening of relatively large mixtures, containing ∼103-106 components. Once the active beads have been identified, the method employed to determine the compounds on those beads must be very sensitive, capable of high throughput, and able to reduce the number of equivocal or redundant identifications to a minimum in order to avoid superfluous resynthesis and testing. Mass spectrometry is the current method of choice for this function. Methods for obtaining mass spectral data for very small amounts of compound on single beads and elucidating the compound identity either directly or indirectly from these data have been demonstrated.2-7 The indirect methods involve either (1) Stankova, M.; Issakova, O.; Sepetov, N. F.; Krchnak, V.; Lam, T. S.; Lebl, M. Drug Dev. Res. 1994, 33, 146.

4354 Analytical Chemistry, Vol. 69, No. 21, November 1, 1997

Table 1. 20 × 20 Dipeptide Model Library, R1-R2-Chelator, and 10 × 10 × 10 Tripeptide Model Library, R1-R2-R3-Chelatora amino acids used for R1 and R2 in dipeptide

amino acids used for R1, R2, and R3 in tripeptide

A R N D C E Q G H I L K M F P S T W Y V

F L W K E H A D Q R

a Composition of the chelator was randomly selected and did not effect outcome of analyses.

tagging the bead at each step of the synthesis or partially capping the compound as it is built up.2,3 These strategies are effective and dependable but may complicate and/or limit flexibility in the chemistry and screening procedures. The direct methods appear to be capable of providing the needed information4-7 and permit greater flexibility in the synthetic and screening steps. However, the specificity of these methods (i.e., their ability to discriminate between candidate structures, thereby reducing ambiguities in the identification of the compound) has not been systematically addressed. Specific analysis strategies, data precision require(2) Geysen, H. M.; Wagner, C. D.; Bodnar, W. M.; Markworth, C. J.; Parke, G. J.; Schoene, F. J.; Wagner, D. S.; Kinder, D. S. Chem. Biol. 1996, 3, 679. (3) Youngquist, R. S.; Fuentes, G. R.; Lacey, M. P.; Keough, T. J. Am. Chem. Soc. 1995, 117, 3900. (4) Egner, B. J.; Langley, G. J.; Bradley, M. J. Org. Chem. 1995, 60, 2652. (5) Zambias, R. A.; Boulton, D. A.; Griffin, P. R. Tetrahedron Lett. 1994, 35, 4283. (6) Hemling, M. E.; Gaitanopoulos, D. E.; Hertzberg, R. P.; Johnson, W. P.; Mentzer, M.; Roberts, G. D.; Taylor, P.; Weinstock, J.; Carr, S. A. Proceedings 43rd Annual Conference Mass Spectrom. Allied Topics, Atlanta, GA, May 2126, 1995. (7) Brummel, C. L.; Vickerman, J. C.; Carr, S. A; Hemling, M. E.; Roberts, G. D.; Johnson, W. P.; Weinstock, J.; Gaitanopoulos, D. E.; Benkovic, S. J.; Winograd, N. Anal. Chem. 1996, 68, 237. S0003-2700(97)00405-8 CCC: $14.00

© 1997 American Chemical Society

Table 2. 12 × 96 × N Nonpeptidic Model Library: Partial Listing of Compositions of Building Blocks R1 Arbitrary Scaffold R3

R2

composition of subunits at R1 code

C

H

N

A1 A2 A3 A4 A5 A6 A7 A8 A9 A10 A11 A12

6 5 6 7 4 4 15 3 5 10 6

13 6 13 8 9 11 16 8 12 14 8 1

2 1 2 1 2 2 1 1 1 1 1

O

S

compositions for R2 F

Cl

Br

1 1 1

1 3 1 1

ments, and the inherent limitations of direct mass spectrometric approaches to structure determination in combinatorial applications have not been established. The limitations and data precision requirements for the mass spectrometric characterization of biopolymers have been studied by computer modeling.8 In the present work, computer models for establishing the specificity of a mass spectrometric identification of compounds from combinatorial libraries are presented; these models are used to examine strategies for obtaining greater specificity in compound identification; the data precision requirements of these analyses and the effects of some library attributes are investigated. MODEL LIBRARIES AND TECHNIQUES The objective was to construct a limited number of model libraries which would demonstrate the general attributes and behaviors of most conceivable libraries. Libraries in which (8) Zuberev, R. A.; Hakansson, P.; Sundqvist, B. Anal. Chem. 1996, 68, 4060.

code

C

H

B13 B14 B15 B16 B17 B18 B19 B20 B21 B22 B23 B24 B25 B26 B27 B28 B29 B30 B31 B32 B33 B34 B35 B36 B37 B38 B39 B40 B41 B42 B43 B44 B45 B46 B47 B48 B49 B50 B51 B52

4 9 7 10 8 6 6 6 12 7 13 8 5 2 8 5 5 15 15 8 8 9 7 5 5 13 5 4 13 11 7 4 6 11 10 8 10 8 8 5

1 6 4 13 9 3 3 5 9 7 10 8 10 3 7 9 9 11 12 7 7 7 11 8 9 8 8 7 8 7 9 7 9 19 10 8 10 5 6 8

N 1 1 2 2

1 1 1

1

1 1 1 1 2

1 1 1 1 1 1

O

S

2 2 2 3 3 2 3 4 2 4 1 1 1 2 2 2 2 2 1 2 3 4 2 2 2 1 1 3 1 2 3 2 2 2 1 2 3 1 2 3

2 2 1 1 1 2 1 2 1 2

F

Cl

Br

1

1

1

3

compounds are built up through repetitive use of a small number of subunits are represented by two peptide models, a 20 × 20 dipeptide and a 10 × 10 × 10 tripeptide (see Table 1). More diverse libraries utilizing larger numbers of building blocks in their construction, typical of nonpeptidic libraries, are represented by a model utilizing three building blocks attached to an arbitrary scaffold (Table 2). The number and diversity of the constituents at R1 and R2 are such that significant changes in the number and/or identities of these subunits does not substantially effect the outcome of the analyses. The number and identities of constituents at R3 were varied to assess the effect of library size and verify the generality of the conclusions. The mass spectral quantities (mass, isotope ratios, masses of MS-MS product ions) for each component of the libraries were calculated and analyzed using Microsoft EXCEL. Several methods were used to analyze the simulated data depending on the information sought. Table 3 shows a portion of the Excel simulation used to determine the specificity obtained from a discrete mass filter and a combined mass/MS-MS filter. The Analytical Chemistry, Vol. 69, No. 21, November 1, 1997

4355

Table 3. Excel Simulations of Mass Filter and Mass/MS-MS Filter A

B

C

D

E

F

G

36 37

mass precision (ppm)

2

38 39

component formula

mass data

MS-MS data

convoluted mass/ MS-MS

mass outcome

mass/MS-MS outcome

40 41 42 43 44 45 46 47 48 49 50 51 52

a71b8 a94b2 a53b4 a67b6 a24b8 a55b8 a72b4 a79b2 a44b4 a27b8 a32b8 a73b12 a39b2

482.1477 483.0702 483.1429 483.1429 483.1429 483.1429 484.0906 484.1270 484.1270 484.1270 484.1270 485.0746 485.1222

133 96 86 105 134 134 87 97 87 135 135 177 98

4 821 477.133 4 830 702.096 4 831 429.086 4 831 429.105 4 831 429.134 4 831 429.134 4 840 906.087 4 841 270.087 4 841 270.097 4 8412 70.135 4 841 270.135 4 850 746.177 4 851 222.098

a

b

degenerate degenerate degenerate degenerate

degenerate degenerate

degenerate degenerate degenerate degenerate

degenerate degenerate

a )IF(OR(AND(C40>C39-C40*$C$36/1000000,C40C41-C40*$C$36/1000000,C40 MS-MS filter > isotope filter. For libraries using less diverse sets of building blocks (exemplified by the two peptide models) the order is MSMS filter > mass filter > isotope filter. Substantially better specificity is achieved through the concerted application of multiple mass spectrometric filters. Multiple filters can be expected to provide unequivocal outcomes for >80% of the components in mixtures of 103-104 components with the remaining degeneracies at the 2-3-fold level. Of perhaps greater importance is the greatly enhanced robustness of the multiple filters. Multiple filter approaches using moderate precision data (e.g., 50 ppm mass precision and 15% isotope ratio precision) can provide greater specificity than discrete filters using very high precision data (e.g., 1 ppm mass precision). The key to achieving greater specificity in structure determinations for large combinatorial mixtures appears to lie in the concerted use of several diverse types of information rather than the application of higher precision data. Future improvements in this area are likely to result from the invention of new mass spectral data filters and the creative application of modeling procedures. For example, modeling can be used in the design of combinatorial libraries to reduce the number of degenerate components in libraries, thereby increasing the specificity of the structure determination. With the recent advances in ion trap and FT/MS technology, high-throughput screening strategies utilizing MSn are a possibility and should be investigated as a way of improving specificity. Finally, this investigation focused on the identification of compounds from resin-bound libraries, but the results clearly are Analytical Chemistry, Vol. 69, No. 21, November 1, 1997

4361

general and apply to the identification of compounds from any combinatorial library of known composition. Further, it should be possible to use similar techniques and develop these principles into noncombinatorial applications in which the composition of the candidate structure pool is not as well defined.

natorial model, and to Steve Brenner for helpful discussions and comments on the manuscript.

ACKNOWLEDGMENT Thanks to Andy Combs for making the tripeptide combinatorial mixture, to Dean Wacker for designing the nonpeptidic combi-

AC970405A

4362

Analytical Chemistry, Vol. 69, No. 21, November 1, 1997

Received for review April 16, 1997. Accepted August 14, 1997.X

X

Abstract published in Advance ACS Abstracts, October 1, 1997.