Ligand-based Virtual Screening

Introduction

The idea of using molecular similarity  to search molecular databases has a long history [1]. Many recent studies have compared shape-matching (ligand-based screening) to docking (ligand-protein screening)  [2, 3, 4]. Docking tools do not address, in adequate detail, the affinity of molecules when compared to a query molecule or a pharmacaphore. The use of shape-based, ligand-oriented methods is therefore justified as an alternative method to determine this functionality.

An overview of InhibOx's in-house tools is given below. Combining the best of these results (super-positional, non-superpositional, electrostatic and feature-based) gives a diverse set of potentially similar ligands for subsequent analysis or assay.

 

Chiral Shape Recognition

InhibOx has developed a non-superpositional in-house method for ligand-based screening called Chiral Shape Recognition (CSR). A key enhancement over previously described non-superpositional methods is that CSR takes into account the chirality of the molecules being compared, while retaining the speed and efficiency of these methods. These differences are important because interactions between proteins and small molecules are often chiral in nature. Using CSR similarly shaped compounds can be quickly identified from within even the largest molecular databases. In addition, the problematic requirement of aligning molecules for comparison is circumvented, as the proposed distributions are independent of molecular orientation. CSR has been demonstrated [5] to provide superior enrichment to previously described methods.

 

FOx

FOx is an in-house software tool for transformation-invariant shape-based screening of a library against a query molecule. It uses a variation of the MaP descriptor method introduced by Steifl and Baumann [6]. The MaP descriptor is based on radial distribution functions (RDF) using distance dependent count statistics. For each pair of atoms, the distance is calculated and the occurrence of the distance is added to the descriptor vector. This distance vector is then used to compare its shape with the shape of the query molecule. Cut-off values for the comparison were manually generated with an exhaustive test over a range of values to determine its optimal value.

The operation of this component is as follows: given a query molecule Q and an evaluation molecule M, FOx initially calculates the array RDF(Q) and RDF(M) using Eq. (1). Here, dij represents the distance between atom i and atom j, and l represents a parameter which denotes the evaluation radius of a particular RDF calculation.

NoA∑tomsNoA∑toms 1 2 RDFl = e2((l+1)-d(i,j)) i=1j=1
 (1)

The mean square RDF (MSRDF) of the two molecules is then calculated using Eq. (2), where RDF(Q) represents the RDF value of the query molecule, whereas RDF(M) represents the RDF value of the evaluation molecule, and L is a user-defined maximum evaluation radius. A user-defined cut-off is then used to filter the MSRDF value to determine shape similarity.

L M SRDF = ∑ (RDF Q - RDF M )2 l=0 l l
 (2)

 

InhibOx’s in-house molecule shape comparison platform has been implemented in an efficient manner to enable the screening of many millions of molecules, represented in 3-D conformational databases, in a reasonable amount of time. Molecules with similar shapes are found by comparing volume overlaps having aligned them onto a common coordinate frame.

The software uses a customised simplex optimiser based method to align database molecules against the query molecule and then assigns scores for their similarity [7] using the Tanimoto(Eq. (1)) and Carbo (Eq. (2)) measures. A good review of similarity measures currently in use can be found in the paper by Raymond and Willett [8].

 

-------C2Q,M-------- TQ,M = C2Q,Q +C2M,M - C2Q,M
 (1)

 

C2Q,M χQ,M = ∘--2---2---- C Q,QC M,M
 (2)

 

Chemical Feature Virtual Screening

InhibOx has developed and validated in-house methods for identifying structural similarity based on recognition of structural features (fingerprint searching).

Fingerprint searching measures similarity using the Tanimoto scoring function (eq. 2, above) and feature detection based on chemical functional groups.

Electrostatic Virtual Screening

InhibOx has developed ElectroShape [9] for ultra-fast comparison of molecules based on the electrostatic properties of the atoms, in addition to the molecular shape and stereochemistry. Combining these two properties maximizes the discovery of relevant lead molecules within the top few percent of structures screened, nearly doubling the enrichment ratio at 1% over previously published shape-based methods, USR and CSR.  While validating the ElectroShape method using release 2 of the Directory of Useful Decoys (DUD) [10], a discrepancy between the partial charges of the active ligands and the decoy ligands was discovered; consistent sets of charges for the DUD molecules have been recalculated, using AMSOL, AM1, Gasteiger and MMFF94-modified methods, and can be downloaded from both the InhibOx and UCSF web sites.

 

 

References

[1] A.C. Good, E.E. Hodgkin, and W.G. Richards (1992). "Similarity screening of molecular data sets." Journal of Computer-Aided Molecular Design, 6(5): 513-520.

[2]   P.C.D. Hawkins, A.G. Skillman, and A. Nicholls (2007). "Comparison of shape-matching and docking as virtual screening tools." Journal of Medicinal Chemistry, 50(1): 74–82.

[3]   R.P. Sheridan, G.B. McGaughey, and W.D. Cornell (2008). "Multiple protein structures and multiple ligands: effects on the apparent goodness of virtual screening results." Journal of Computer-Aided Molecular Design, 3-4: 257–265.

[4]   J. Venhorst, S. Núñez, J. W. Terpstra, and C.G. Kruse (2008). "Assessment of scaffold hopping efficiency by use of molecular interaction fingerprints." Journal of Medicinal Chemistry, 51(11): 3222-3229.

[5] M.S. Armstrong, G.M. Morris, P.W. Finn, R. Sharma, and W.G. Richards (2009). "Molecular similarity including chirality", Journal of Molecular Graphics and Modelling, 28: 368-370.

[6]   N. Stiefl and K. Baumann (2003). "Mapping property distributions of molecular surfaces: Algorithm and evaluation of a novel 3D quantitative structure-activity relationship technique." Journal of Medicinal Chemistry, 46(8): 1390–1407.

[7]   P. Willett, J.M. Barnard, and G.M. Downs (1998). "Chemical similarity searching". Journal of Chemical Information and Computer Sciences, 38(6): 983–996.

[8]   J.W. Raymond and P. Willet (2002). "Effectiveness of graph-based and fingerprint-based similarity measures for virtual screening of 2D chemical structure databases." Journal of Computer-Aided Molecular Design, 16: 59–71.

[9]    M.S., Armstrong, G.M., Morris, P.W., Finn, R. Sharma, L. Moretti, R.I. Cooper, and W.G. Richards (2010). "ElectroShape: fast molecular similarity calculations incorporating shape, chirality and electrostatics." Journal of Computer-Aided Molecular Design 24(9): 789–801. Epub 2010 Jul 8.

[10]    N. Huang, B.K. Shoichet, and J.J. Irwin (2006). "Benchmarking Sets for Molecular Docking" Journal of Medicinal Chemistry, 49(23): 6789–6801.