Encoding Protein–Ligand Interaction Patterns in Fingerprints and Graphs

Abstract

We herewith present a novel and universal method to convert protein-ligand coordinates into a simple fingerprint of 210 integers registering the corresponding molecular interaction pattern. Each interaction (hydrophobic, aromatic, hydrogen bond, ionic bond, metal complexation) is detected on the fly and physically described by a pseudoatom centered either on the interacting ligand atom, the interacting protein atom, or the geometric center of both interacting atoms. Counting all possible triplets of interaction pseudoatoms within six distance ranges, and pruning the full integer vector to keep the most frequent triplets enables the definition of a simple (210 integers) and coordinate frame-invariant interaction pattern descriptor (TIFP) that can be applied to compare any pair of protein-ligand complexes. TIFP fingerprints have been calculated for ca. 10,000 druggable protein-ligand complexes therefore enabling a wide comparison of relationships between interaction pattern similarity and ligand or binding site pairwise similarity. We notably show that interaction pattern similarity strongly depends on binding site similarity. In addition to the TIFP fingerprint which registers intermolecular interactions between a ligand and its target protein, we developed two tools (Ishape, Grim) to align protein-ligand complexes from their interaction patterns. Ishape is based on the overlap of interaction pseudoatoms using a smooth Gaussian function, whereas Grim utilizes a standard clique detection algorithm to match interaction pattern graphs. Both tools are complementary and enable protein-ligand complex alignments capitalizing on both global and local pattern similarities. The new fingerprint and companion alignment tools have been successfully used in three scenarios: (i) interaction-biased alignment of protein-ligand complexes, (ii) postprocessing docking poses according to known interaction patterns for a particular target, and (iii) virtual screening for bioisosteric scaffolds sharing similar interaction patterns.

Binary Quantitative Structure-Activity Relationship Analysis in Retrospective Structure-Based Virtual Screening Campaigns Targeting Estrogen Receptor Alpha

“…Similar to "bad" identified compounds by SPORES1.3, screened compounds that could not result in docking pose in this step were tagged as in actives or negatives (N). The enrichment factor (EF) [26,31] and F-measure [2,31] value calculations were adjusted by considering the "bad" identified compounds by SPORES1.3 and the failed screened compounds as in actives or negatives (N). Ligands predicted as actives or positives (P) were encoded as true positives (TP), while ligands predicted as N were then encoded as false negatives (FN).…”

“…The IFP which is also known as the proteinligand IFP (PLIF) has been successfully employed mainly in fragmentbased drug discovery projects [1][2][3][4][5][6]. Inspired from IFP of Marcou and Rognan, an open-source Python implementation of the molecular IFP named PyPLIF was developed [7,8].…”

Binary Quantitative Structure-Activity Relationship Analysis to Increase the Predictive Ability of Structure-Based Virtual Screening Campaigns Targeting Cyclooxygenase-2

“…The development of methods and computer applications to identify and compare Protein-Ligand Interaction Fingerprints (PLIF) [1][2][3][4][5] and its variances has been of considerable interest since employing such fingerprints was a promising strategy to leverage the wealth of generated data in rational drug design [6]. Together with Structural Interaction Fingerprint (SIFt) and molecular interaction fingerprinting [6], PLIF was categorized as binding site-focused interaction fingerprinting methods [2].…”

“…Inspired by this molecular interaction fingerprinting [1], our research group developed a Python implementation of PLIF featuring hydrogen bonds, ionic and hydrophobic interactions and π-stacking under the name PyPLIF [4][5]. Molecular interaction fingerprinting [1] has been expanded recently by encoding the patterns of protein−ligand interactions in fingerprints and graphs that could be applied for post-processing docking poses and search for plausible bioisosteric scaffolds [2][3].…”

“…Rescoring the results of the molecular docking simulations by calculating Tanimoto-coefficient similarity with a PLIF reference (Tc-PLIF) has been shown to increase the predictive ability of several SBVS campaigns [1,6,[11][12][13] and to better re-dock small molecules in their native poses [1,4,6] compared to standard docking scores [2][3]. Notably, different PLIF references could lead to very different paths of the discoveries [12,14].…”