Statistical tools for virtual screening

Journal of Medicinal Chemistry

Volumn 48, Issue 23, 2005, Pages 7477-7481

  Krumrine, Jennifer R ^a   Maynard, Andrew T ^a,b   Lerman, Charles L ^a  

^a ASTRAZENECA   (United Kingdom)

^b Icagen Inc   (United States)

Author keywords

[No Author keywords available]

Indexed keywords

ARTICLE; ASSAY; DATA BASE; SAMPLING; SCORING SYSTEM; SIGNAL NOISE RATIO; STATISTICAL ANALYSIS; STATISTICAL MODEL; VIRTUAL REALITY;

ALGORITHMS; BINDING SITES; DATABASES, FACTUAL; LIGANDS; MODELS, MOLECULAR; PHARMACEUTICAL PREPARATIONS; PROBABILITY; PROTEINS; STATISTICS;

EID: 28544434157     PISSN: 00222623     EISSN: None     Source Type: Journal    
DOI: 10.1021/jm0501026     Document Type: Article

References (16)

1. Bissantz, C.; Folkers G.; Rognan D. Protein-based Virtual Screening of Chemical Databases. 1. Evaluation of Different Docking/Scoring Combinations. J. Med. Chem. 2000, 43, 4759-4767.
2. Lyne, P. Structure-Based Virtual Screening: An Overview. Drug Discovery Today 2002, 7, 1047-1055.
3. Friesner, R. A.; Banks, J. L.; Murphy, R. B.; Halgren, T. A.; Klicic, J. J.; Mainz, D. T.; Repasky, M. P.; Knoll, E. H.; Shelley, M.; Perry, J. K.; Shaw, D. E.; Francis, P.; Shenkin, P. S. Glide: A New Approach for Rapid, Accurate Docking and Scoring. 1. Method and Assessment of Docking Accuracy. J. Med. Chem. 2004, 47, 1739-1749.
4. 5 for the appropriate generation of ionization, tautomeric, and configurational states. All isomers were docked independently, and the best score was retained as the Glide score for the compound.
5. Kenny, P. W.; Sadowski, J. Structure Modification in Chemical Databases. Chemoinformatics in Drug Discovery; Oprea, T. I., Ed.; Wiley-VCH Verlag GmbH & Co. KGaA: Weinheim, Germany, 2005; Vol. 23, pp 271-284.
6. Taylor, R. Simulation Analysis of Experimental Design Strategies for Screening Random Compounds as Potential New Drugs and Agrochemicals. J. Chem. Inf. Comput. Sci. 1995, 35, 59-67.
7. 2] = [0,X], and for X ≥ N·P, the interval is [X,N], where X is the number of compounds in the neighborhood scoring above a designated threshold, N is the neighborhood size, and P is the hit rate of "actives" in the virtual screen. The sign of the score distinguishes true (+) and false positive (-) domains. N_score > 1.3 indicates a compound neighborhood with a statistically significant number of hits (95% confidence). In MATLAB, N_score = -log 10(min(binocdf-(X,N,P),1-binocdf(X-1,N,P))) sign(X - N·P). Equivalently, N_score can be evaluated using binopdf in a loop, which avoids a potential roundoff error in the binocdf function.
8. Daylight Chemical Information Systems, Inc.: Irvine, CA; http://www.daylight.com.
9. Willett, P.; Barnard, J. M.; Downs, G. M. Chemical Similarity Searching. J. Chem. Inf. Comput. Sci. 1998, 38, 983-996.
10. Butina, D. Unsupervised Data Base Clustering Based on Day-light's Fingerprint and Tanimoto Similarity: A Fast and Automated Way To Cluster Small and Large Data Sets. J. Chem. Inf. Comput. Sci. 1999, 39, 747-750.
11. As an alternative (unpublished) to clustering a database, the neighborhood of each compound in a database is sampled locally and independently. For example, the local neighborhood of a given compound is defined as the set of compounds within a specified Tanimoto radius (e.g., 0.3 units) of the compound. In this manner, neighborhoods are locally sampled for each compound and N_score is computed according to eq 3. Clusters are neither computed nor stored, avoiding potential difficulties in assigning the cluster membership of compounds.
12. Martin, Y. C.; Kofron, J. L.; Traphagen, L. M. Do Structurally Similar Molecules Have Similar Biological Activity? J. Med. Chem. 2002, 45, 4350-4358.
13. Lipinski, C. A. Drug-like properties and the causes of poor solubility and poor permeability. J. Pharmacol. Toxicol. Methods 2000, 44, 235-249.
14. Mandel, J. The Statistical Analysis of Experimental Data; Dover Publications: New York, 1964.
15. One could decide that acceptable risk limits are neighborhood-specific; for example, if a particular neighborhood contains a large number of compounds that are readily available for testing compared to other neighborhoods that would require extensive synthesis, it may be worthwhile to adopt a sampling plan for neighborhood-specific sampling that is more computationally expensive but provides more information.
16. We believe that by eliminating those few compounds in any given neighborhood that clash with the receptor or find no reasonable pose, possibly due to inadequate sampling, one derives the clearest picture of the neighborhood behavior; otherwise, one compound with a score of 10 000 would disallow selection of that particular neighborhood, even if all other compounds have excellent dock scores. Further, one might make an exception to prefilter compounds with a very large number of rotatable bonds from the docking database, which can be extremely time-consuming to dock, especially if the neighborhood is reasonably well represented by similar compounds with fewer rotatable bonds.