Quantitative surface field analysis: learning causal models to predict ligand binding affinity and pose

Abstract

We introduce the QuanSA method for inducing physically meaningful field-based models of ligand binding pockets based on structure-activity data alone. The method is closely related to the QMOD approach, substituting a learned scoring field for a pocket constructed of molecular fragments. The problem of mutual ligand alignment is addressed in a general way, and optimal model parameters and ligand poses are identified through multiple-instance machine learning. We provide algorithmic details along with performance results on sixteen structure-activity data sets covering many pharmaceutically relevant targets. In particular, we show how models initially induced from small data sets can extrapolatively identify potent new ligands with novel underlying scaffolds with very high specificity. Further, we show that combining predictions from QuanSA models with those from physics-based simulation approaches is synergistic. QuanSA predictions yield binding affinities, explicit estimates of ligand strain, associated ligand pose families, and estimates of structural novelty and confidence. The method is applicable for fine-grained lead optimization as well as potent new lead identification.

Design, Synthesis, and Evaluation of Novel 2-Methoxyestradiol Derivatives as Apoptotic Inducers through an Intrinsic Apoptosis Pathway

“…The 3D QSAR study was generated by using 20 molecules from synthesis, 2-methoxyestradiol (1) and their negative logarithm of the IC 50 value (pIC 50 ) of MCF-7, performed with Surflex (Version 4.4) using leave-one-out cross validation method as previously described [41]. Twenty randomly selected molecules were calculated by Surflex-Quansa to build a 3D QSAR model (number of molecules to select for core multiple-alignment = 10) by leaving a single molecule out.…”

“…A 3D QSAR study was generated by using these 21 molecules and their pIC 50 value of MCF-7, performed with Surflex-Quansa using the leave-one-out cross validation method as previously described [41]. The molecules with low activity (pIC 50 of MCF-7 < 4) were in correlation with predicted values.…”

Leveraging non-structural data to predict structures of protein–ligand complexes

“…Likewise, ComBind can be used with any pairwise pose similarity metric or combination thereof. ComBind's performance could potentially be improved by using more fine-grained interaction descriptors (41,42) or by using similarity metrics based on field-based methods developed for virtual screening (28,43).…”

Electrostatic-field and surface-shape similarity for virtual screening and pose prediction

“…Here, we introduce a new 3D similarity method that combines the surface-based approach of Surflex-Sim and related methods [14–16] with an electrostatic field comparison method that derives from the recently introduced QuanSA 3D-QSAR approach [17]. The method is called “eSim” for e lectrostatic-field and s urface- s hape sim ilarity (with the three “s” characters giving rise to the capital “S”).…”

“…Figure 1 depicts the central calculation within eSim, which is the computation of feature values at observer points, with the latter having been placed outside of a query molecule (also referred to as a “target molecule” in what follows). At each observer point, six values are computed:: The distance in Angstroms from the point to the atom surface, which corresponds to the minimum over the distances to each atom less that atom’s VdW radius.: The molecular electric field is characterized in terms of the Coulombic energy of moving a point charge of + 0.2 e to the observer point from an infinite distance (as in the QuanSA approach [17], with additional details provided in the Methods Section).: The minimum distance from the observer to any donor proton’s surface.: The angle formed by the observer point, the donor proton, and the atom attached to the donor proton.: The minimum distance from the observer to any acceptor atom’s surface.: The angle formed by the observer point, the acceptor atom, and the centroid of the atoms attached to the acceptor atom (180°, for example, when a pyridine nitrogen is oriented perfectly toward an observer).

Fig. 2Five small molecules (cyan) are shown in their optimal poses relative to the query/target 2-pyrrolidone (magenta, shown with yellow observer points, upper left), with their corresponding eSim scores

…”