Plane of best fit: a novel method to characterize the three-dimensionality of molecules

Abstract

We describe a computational method, plane of best fit (PBF), to quantify and characterize the 3D character of molecules. This method is rapid and amenable to analysis of large diverse data sets. We compare PBF with alternative literature methods used to assess 3D character and apply the method to diverse data sets of fragment-like, drug-like, and natural product compound libraries. We show that exemplar fragment libraries underexploit the potential of 3D character in fragment-like chemical space and that drug-like molecules in the libraries examined are predominantly 2D in character.

Figure 1

Structures of protein–protein interaction inhibitors with pronounced 3D shape in the respective protein–ligand X-ray crystal structures: (a) (+)-JQ1, an inhibitor of bromodomain BRD4/chromatin interactions, (b) inhibitor of LEDGF/p75 integrase interaction, (c) Nutlin-3, an inhibitor of Bcl2/BH3 domain interactions, and (d) ABT737, an inhibitor of MDM2/p53 interactions.

Figure 2

Example depiction of plane of best fit (PBF) for cyclohexane in a chair conformation.

Figure 3

PMI plot for the ICRFL data set. Four evenly distributed bins across the range of the average distance from the plane of best fit (PBF score) are represented by the color of the points (red ≤ 0.2709, 0.2709 < blue ≤ 0.5418, 0.5418 < yellow ≤ 0.8128, black > 0.8128).

Figure 4

Scatter graph measuring the correlation between molecular globularity (MolGlob) and PBF score for the ICR fragment library data set.

Figure 5

Scatter graph depicting the correlation between fraction of sp³ hybridized carbons (Fsp³) and PBF score for the ICR fragment library data set.

Figure 6

Chemical structures and CORINA-derived conformation comparison for molecules (1) and (2) for which Fsp³ = 0.

Figure 7

Chemical structures and CORINA-derived conformation comparison for molecules (3) (Fsp³ = 0.476) and (4) (Fsp³ = 0.474).

Figure 8

Example molecules selected from the ICRFL data set (5–11) and their respective PBF scores depicted on a linear scale.

Figure 9

Shape comparison of molecules (12) and (13) which score similarly according to literature methods.

Figure 10

Density plot of PBF score versus the sum of normalized principal moments of inertia (NPR) for the eMolecules data set with acyclic and Ro5 noncompliant compounds removed. The horizontal black line represents a cutoff for 3D molecules for NPR1 + NPR2, and the vertical line, a corresponding cutoff for PBF.

Figure 11

Range of PBF scores depicted as box and whisker plots across all nine data sets. Yellow plots include all members of the data sets, red plots are after a Lipinski Ro5 filter has been applied.

Figure 12

Histogram of PBF scores of ICR and Maybridge fragment libraries, the GDB-13 data set, DrugBank, and ChEMBL libraries subject to RECAP algorithm, and application of fragment-like filters.