Journal Articles

Overview of Pharmacophore Applications in MOE

Anna Lin
Chemical Computing Group Inc.

Introduction | Pharmacophores in MOE | Data | Generating a Query | Pharmacophore Search | Consensus Query | Volumes | Summary | References



Understanding the chemical and structural details of receptor active sites and receptor-ligand interactions is fundamental to the successful search for candidate active ligands. An important technique for developing insight into which features are important for biological activity is the encoding of chemical structural features or pharmacophores into a 3D query that can be matched against molecular data. With such queries, it is possible to test pharmacophoric hypotheses, do searches over databases of candidate ligands, generate new hypotheses, and perform iterative query refinement.

This article overviews the pharmacophore applications in Chemical Computing Group's Molecular Operating Environment (MOE). These tools provide interactive controls for generating 3D queries and for doing database searches. For purposes of illustration, we will demonstrate how a query can be easily generated and refined using a test database of 38 ligands, and then apply the query to a second database of 1957 compounds to identify new candidate molecules.

Pharmacophores in MOE

MOE's pharmacophore applications use a general notion of a pharmacophore:

A pharmacophore is a set of structural features in a ligand that are directly related to the ligand's recognition at a receptor site and its biological activity.

In MOE, pharmacophoric structural features are represented by labeled points in space. Each ligand is assigned an annotation, which is an encoding of the structural features in the ligand that may contribute to the ligand's pharmacophore. Which structural features are encoded is determined by the currently selected pharmacophore scheme. For example, the Polar-Charged-Hydrophobic scheme assigns the label Aro to an aromatic center, while the Planar-Polar-Charged-Hydrophobic scheme would assign to the same structure the label HydP, indicating a planar hydrophobic area.

A database of annotated ligands can be searched with a query that represents a pharmacophore hypothesis. A query is a collection of feature, feature constraint, and volume restrictions that is applied to the annotation and atoms of a ligand conformation. The result of such a database search is a set of ligand conformations for which all restrictions of the query are satisfied. Partial matches, where only a subset of the restrictions is met, are also possible.


For this article we used two databases. A database of compounds comprising known active and inactive 17beta-hydroxysteroid dehydrogenase type 1 (17beta-HSD-Type1 enzyme) inhibitors [Makela 1995, 1998; Hoffren 2001] was used to create the pharmacophore query. The composition of this database was 7 steroidal compounds, 21 flavonoid, and 10 non-steroidal and non-flavonoid for a total of 38 molecules. Note that although activity data was not available for estradiol and equilin, both were also included in the test set.

The pharmacophore was then applied to a second database (the "search" database) of 1957 compounds of unknown inhibition activity to search for possible new candidates.

The 17beta-HSD-Type1 enzyme catalyzes the conversion of estrone to 17beta-estradiol. In breast cancer tumors, the concentration of estrogen has been found to exceed that of normal breast tissue, thus inhibition of the enzyme may promote regression of such tumors by reducing tumor estrogen levels [Poutanen 1995].

Two kinds of activity data were available for compounds in the test database: the inhibition percentage of conversion of estrone to estradiol with the purified 17b-HSD-Type1 enzyme, and the inhibition percentage of conversion with the T-47D breast cancer cell. In both cases, we used activity values for ligands tested at 1.2 microM concentration. Three inhibitors of particular interest for their strong inhibitory properties are apigenin, a flavone, genistein, an isoflavone, and coumesterol, a non-flavonoid. Of all the test compounds, apigenin is the best inhibitor in the cancerous biological system.

Preparing the Data

The 38 ligands of the test database were built in MOE, minimized, and conformers then generated using MOE's Conformation Import tool, for a resultant database of 649 conformers. Structural data of the enzyme, complexed with estradiol and the NADPH co-factor (1FDT from the Protein Data Bank [Berman 2000]), was used to direct the preparation of both ligands and pharmacophore queries.

The conformers of the 1957 molecules of the search database were also generated (using MOE/smp distributed computing technology), for a total of just over 155000 conformers.

Generating a Query

In this section, we show how to construct two pharmacophore queries. The first derives from the original complexed molecule and is meant to be a general query for molecules that show a similar binding pattern to the native ligand. The second is one specific for flavonoids.

For generating an initial pharmacophore query, we begin by examining the original complexed system, 1FDT. In the crystal structure, the substrate is in hydrogen bond contact with the residues TYR155, SER142, and HIS221. TYR155 and SER142 are both in contact with the same oxygen atom on the ligand. The interaction with TYR155 is essential to the biochemical reduction reaction catalyzed by the enzyme, allowing for hydrogen transfer between the co-factor (NADPH) and the substrate. The other two hydrogen bonds are important for substrate positioning.

Cutaway View of 1FDT Pocket with Estradiol Bound
Cutaway View of 1FDT Pocket with Estradiol Bound

Estradiol was used as a starting point for generating a pharmacophoric query. MOE's Pharmacophore Query Editor automatically detects pharmacophoric features and generates annotations. With the estradiol molecule loaded in the system, a four-feature pharmacophore was quickly generated.

Estradiol Annotation
Estradiol with Pharmacophore Annotation
Estradiol Query
Estradiol with Pharmacophore Query

Two hydrophobic features were created at the junctions of rings A and B and of rings C and D by selecting two ring annotation points at a time and then pressing Feature in the Query Editor. Feature F2, where the interaction with HIS221 takes place, was edited to be an acceptor instead of a donor-acceptor. Both F1 and F2 were marked as being essential.

To construct a query that is selective for flavonoids, we began with flavone (note, however, that four of the flavonoids in the test database were isoflavones). From the four pharmacophoric annotation points, three were chosen to construct a 3-point pharmacophore, as illustrated below.

Flavone Query
Flavone Query

This query differs from the estradiol query in several significant ways. Although the two aromatic features overlap the hydrophobic features of the estradiol query, they are smaller, and both are specifically aromatic. Also, the hydrogen bond site features of estradiol were not included in the initial flavone query; instead, an acceptor at the carbonyl oxygen was used.

Pharmacophore Search

After generating an initial query, the test database was searched. The results of the search were then examined to suggest modifications to the query.

Prior to searching, pharmacophore annotations were calculated for all molecules in the test database using the Pharmacophore Preprocessor, a utility that is also integrated into the Pharmacophore Search application. This step was done to allow for faster searching.

In the Pharmacophore Search panel, Hit Entries were specified to be selected in the Database Viewer, in addition to being written out to a separate results database. Using the estradiol pharmacophore from above resulted in 4 hit molecules, of which 3 were steroids (including estradiol), and 1 was a flavonoid.

The small hit set suggested that the query might be overconstrained. When a partial match of a minimum of 3 features was permitted, the result was 14 hit molecules, of which 3 were steroidal and 7, flavonoid. The hit set included both estradiol and equilin, as well as coumesterol which, although non-steroidic, is similar to estradiol in structure. About 1/3 of the compounds had good to moderate activity.

Quick experimentation using the Ignore flag in the Pharmacophore Query Editor showed that hydrophobic feature F3 could be eliminated from the query without affecting results.

When this query was applied to the search database, the hit set included most of the steroidal compounds in the database.

Database Search Results for Estradiol Query

With the flavone pharmacophore query, 20 hit molecules were obtained, of which 1 was non-flavonoid (coumesterol), and 19 were flavonoids; none were steroids. Formonetin and daidzein, both isoflavones, were missed by the query. Note, however, that genistein and biochanin A, also isoflavones, were successfully matched. An examination of the output hit conformer database confirmed that the presence of extra hydroxyl groups in these latter compounds permit a good superposition with the pharmacophore query, as shown below for genistein (magenta), superimposed upon flavone (blue). This result underlies the importance of close examination of pharmacophore search results, and may even suggest alternate ligand binding modes.

Note that preference for flavones over isoflavones may in fact be desirable since isoflavones appear to exhibit greater estrogenicity than flavones.

Flavone and Genistein with Flavone Query
Genistein (magenta) Superimposed upon Flavone (blue)

When this flavonoid query was applied to the search database, 350 hits were obtained, or slightly over 18% of the compounds, none of which were steroidal in nature.

Consensus Query

The flavone query is missing the two important hydrogen bonding sites observed in the native bonding interaction of 1FDT. Adding these features to the query may help enforce a better binding position of the candidate ligands. To recover these features, we loaded flavone along with estradiol, apigenin, and coumesterol molecules into MOE. These latter three ligands all exhibit hydrogen bonding contact with TYR155 and HIS221. Three molecules were used preferentially over a single molecule to include some variability in positioning. Using the Pharmacophore Consensus application, we obtained a consensus query. The Pharmacophore Consensus application calculates a query composed of all features of all molecules, with an indication of the proportion of shared features.

Flavone, Estradiol, Apigenin, and Coumesterol Consensus Query in MOE

Flavone (magenta), Estradiol (blue), Apigenin (red), and Coumesterol (green) with Consensus Query
Flavone, Estradiol, Apigenin, and Coumesterol Consensus Query in Panel

We selected the two hydrogen bonding sites, G3 and G4, to be added to the original flavone query. Then, the constraints on the query were relaxed slightly to allow a partial match of a minimum of four features, to reflect the variability in the test ligands. Finally, we relaxed the aromatic condition on the two aromatic features, allowing a non-aromatic hydrophobic feature to also match at those points (the feature type was set to Aro|Hyd).

Flavone, Estradiol, and Coumesterol Consensus Query in MOE
Flavone (magenta), Estradiol (blue), Apigenin (red), and Coumesterol (green) with New Query

A search of the test database using this new consensus query yielded 17 hits, including 15 flavonoids, 1 non-flavonoid (coumesterol), and 1 steroid (estradiol). The hit set included those test compounds having highest inhibitory activity, including coumesterol, apigenin, apigenin analogs, and genistein.

The query was then applied to the search database compounds, in which case 201 hits were obtained, or about 10.5% of the compounds.


A visual inspection of the output of the search revealed molecules that should be ruled out for steric reasons. An additional volume constraint was therefore added to the query to help further constrain the search. Using the Union feature of the Pharmacophore Query Editor, excluded volume spheres were positioned coincident with atoms in the active site of the 1FDT complex that were within 5 A of the bound ligand. The radii of the spheres was set to 1.4 A.

Query with Excluded Volumes
Cutaway View of 1FDT Pocket with Query Containing Excluded Volumes

Using the supplemented query on the search database yielded 134 hits, or about 7% of the molecules in the database.


3D pharmacophore generation and searching is an important technique used in the identification of candidate active ligands. In MOE, 3D queries can contain locations of pharmacophore features or chemical groups as well as restrictions on shape imposed by specifying included and/or excluded volumes. An interactive editor allows for both query customization and elucidation of a consensus query from a set of aligned molecules. Such a query can then be used to filter a conformation database. The pharmacophoric database search application provides a high degree of control, offering both partial and systematic matching as well as flexible matching rules. MOE's pharmacophore tools are integrated together, making the process of iterative pharmacophore model generation and refinement easier.


[Berman 2000] Berman, H.M., Westbrook, J., Feng, Z., Gilliland, G., Bhat, T.N., Weissig, H., Shindyalov, I.N., Bourne, P.E. The Protein Data Bank. Nucleic Acids Research. 28, 235-242 (2000).
[Hoffren 2001] Hoffren, A.-M., Murray, C.M., Hoffman, R.D. Structure-Based Focusing Using Pharmacophores Derived from the Active Site of 17beta-hydroxysteroid dehydrogenase. Current Pharmaceutical Design. 7, 547-566 (2001).
[Makela 1995] Makela, S., Poutanen, M., Lehtimaki, J., Kostian, M.L., Santti, R., Vihko, R. Estrogen-specific 17beta-hydroxysteroid oxidoreductase type 1 (E.C. as a Possible Target for the Action of Phytoestrogens. Proceedings of the Society for Experimental Biology and Medicine. 208, 51-59 (1995).
[Makela 1998] Makela, S., Poutanen, M., Kostian, M.L., Lehtimaki, J., Strauss, L., Santti, R., Vihko, R. Inhibition of 17beta-hydroxysteroid oxidoreductase by Flavonoids in Breast and Prostate Cancer Cells. Proceedings of the Society for Experimental Biology and Medicine. 217, 310-316 (1998).
[Poutanen 1995] Poutanen M, Isomaa V, Peltoketo H, Vihko R. Regulation of Oestrogen Action: Role of 17beta-hydroxysteroid dehydrogenases. Ann. Med. 27, 675-682 (1995).