SELECTION OF PHARMACOPHORICALLY DIVERSE ACIDS FOR PEPTIDO-MIMETIC TYPE LIBRARIES

SELECTION OF PHARMACOPHORICALLY DIVERSE ACIDS

C. Luttmann, S. Pickett , V. Guerin, F-F. Clerc,H. Dubois, A. Laoui and E. James-Surcouf
New Lead Generation Chemistry Department, Molecular Modelling & Protein Crystallography
Medicinal Chemistry Department, Rhone-Poulenc Rorer S.A., CRVA, 13 quai Jules Guesde, B.P.14, 94403 Vitry-sur-Seine, France.

INTRODUCTION

The last few years have seen an explosive growth in the use of combinatorial methods for the creation of extremely large libraries of structurally diverse molecules, from which it has proved possible to identify biologically active molecules far more rapidly than is possible using conventional approaches to drug discovery. The effectiveness of the approach is crucially dependent on the diversity of the reactantsthat are used as the input to the combinatorial synthesis of the final products, since there are generally far more reactants available than can actually be used in practice.

In this poster we will present a strategy developed (1) to define a set of diverse acids as reactants for the generation of peptido-mimetic type libraries based on amide chemistry;this latter having the format R1-(C=3DO)-{NH-Core-C(=3DO)}-NH-R2 (R1 : Rgroup of acids ; R2 : Rgroups of 48 native and non-natural alpha-amino-acids ; Core : amino-carboxylic scaffold).

STRATEGY AND METHODS

Pharmacophore concept & key definition :
Global strategy used
- Extraction of potential reagents from the ACD, filter the list (MW, supplier, toxics, ...)
- Selection of a representative subset of amino-acids at R2
- Creation of virtual library and calculation of a pharmacophore key per Rgroup/reagent representing the final products
- 'Chemical/structural' complexity classification of acids R1
- Stepwise selection of acids (R1) starting with the least complex set
Dissimilarity based reagent selection (DBRS)
- Calculate sum of similarities for each reagent with respect to all others
- Select the most dissimilar one (lowest sum) & place it in the selection subset
- Search for the Rgroup most dissimilar to the ones already selected & place it in the selection set
- Repeat previous step until goal achieved

RESULTS

Selection of a subset of a-amino-acids at R2
Using several peptido-mimetic core-structures, a representative subset of R2 amino-acids has been selected based on their pharmacophore diversity using the DBRS algorithm described above. 11 R2 are sufficient to explore over 90 % of the pharmacophore space i.e., D-Tic, L-Asp, D-His, D-Arg, D-Glu, L-Thr, D-Tyr, D-Trp, D-Lys, and 2 non-natural amino-acids LB, LU.
Complexity analysis of acids
- Divide acids into groups according to their chemical/structural complexity
- Use of Daylight tools (2,5) to analyse structure i.e. 'fingerprint' molecule according to the pharmacophore centers and structural features it contains
- Use scripts (5) to analyse the fingerprints and perform subset selection according to user-defined criteria e.g. max. acceptors, max. rings , max. branched systems, etc ...
Stepwise selection of 90 diverse acids
- A representative core structure of low structural & chemical complexity (no stereo center, low flexibilty) has been used to built a library to be used for the initial selection
- For each acid a pharmacophore key is generated that takes into account all 11 combinations at R2
- Begin selection with the acids the least complex chemically/structurally & least promiscuous pharmacophorically (see figure below)
- Add more complex acids only if compounds add significantly to pharmacophore space covered 3D> Starting from a set of 1544 acids, the final selection of 90 acids covers over 85 % of the pharmacophore space considered. Several other scaffolds have been used to check consistency of this final selection.

A: 1100 acids are classified into different sets cntgrp1, cntgrp2 etc. of increasing structural complexity. cntgrp1 is further divided into two based on the number of pharmacophores (np) exhibited by the final library products.

B: Any previously selected are added to the working set to constitute the starting list.

C: a diverse set is selected using DIVSEL. The two numbers at the bottom of each column give the number of pharmacophores in the selected set and those in all molecules considered to date respectively.

DISCUSSION

Our selection was based on pharmacophore descriptors as implemented in ChemDiverse (3) and using in-house definitions (4). Due to the large number of combinations possible (R1xR2) we have defined a subset of 11 representative alpha-amino-acids at R2 that covers over 90% of the diversity space in that position. This reduces substantially the CPU time needed for the subsequent calculations. The DBRS selection algorithm has the pitfall that it points quite rapidly to complex structures as these exhibit substantially more unique pharmacophores, thus adding more to the diversity space considered, but are of less interest in an initial screening set. This led us to develop procedures to identify less complex structures initially and perform reagent selection in a stepwise manner. The 90 acids so selected cover over 85 % of the diversity space considered. Such a selection has proven to be useful to the chemist for final reagent selection.

Although a pharmacophore intrinsically codes for some physico-chemical properties we are currently evaluating methods that would include such descriptors explicitly in combination with the pharmacophore descriptors. Our current studies involve the use of software such as DiverseSolutions (6).

CONCLUSION

We have defined a set of diverse acids for peptido-mimetic type libraries using pharmacophore-based descriptors combined with a new algorithm for dissimilarity based compound selection and procedures to identify the least complex but most diverse structures.

References :

S.D. Pickett, C. Luttmann, V. Guerin,A. Laoui and E. James, manuscript submitted to JCICS.
Daylight Chemical Information Systems Inc, Los Alto, California, USA.
ChemDiverse, Chemical Design Ltd., Shipping Norton, Oxfordshire,UK.
S. Pickett et al., JCICS, 1996, 36, 6, 1214-1223.
In-house developed program/procedures
R. Pearlman, University of Texas, Austin, USA.

Acknowlegments : We are grateful to J.S. Mason (CVRC) for ChemDiverse procedures and specific parameter files used.