Computational Methodologies for Compound Database Searching that Utilize Experimental Protein–Ligand Interaction Information

Abstract

Ligand-and target structure-based methods are widely used in virtual screening, but there is currently no methodology available that fully integrates these different approaches. Herein, we provide an overview of various attempts that have been made to combine ligand-and structurebased computational screening methods. We then review different types of approaches that utilize protein-ligand interaction information for database screening and filtering. Interaction-based approaches make use of a variety of methodological concepts including pharmacophore modeling and direct or indirect encoding of protein-ligand interactions in fingerprint formats. These interaction-based methods have been successfully applied to tackle different tasks related to virtual screening including postprocessing of docking poses, prioritization of binding modes, selectivity analysis, or similarity searching. Furthermore, we discuss the recently developed interacting fragment approach that indirectly incorporates 3D interaction information into 2D similarity searching and bridges between ligand-and structurebased methods.Key words: fingerprints, interacting fragments, ligand-and structure-based virtual screening, pharmacophores, protein-ligand interactions, similarity searching, X-ray structures Virtual screening has become an important component of drug discovery and medicinal chemistry research. A wide range of computational methods are available for mining of chemical and biologic data to identify novel compounds with desired properties (1,2). In principle, virtual screening approaches are either ligand-based (2-6) or target structure-based (5-10). Small molecule-based methods typically rely on the concept of 'molecular similarity'. An important component of the molecular similarity concept is the Similarity Property Principle (11), which simply states similar molecules should have similar biologic activities. This principle provides the basis for whole-molecule similarity searching for which many different methods have been developed (2,4). Structure-based virtual screening (molecular docking) methods rely on the knowledge of 3D target structure and evaluate energetic and geometric criteria that determine small molecule binding (7-9). Ligand-and structure-based virtual screening approaches utilize different types of molecular information and are often complementary in nature. Accordingly, a number of efforts have been made to combine these approaches (12-18) and exploit their complementarity. Despite these efforts, fully integrated ligand-and structure-based virtual screening systems are yet to be developed. However, in fast-growing public domain databases abundant target and ligand information is becoming readily available. In particular, the Protein Data Bank (PDB) (19) offers the opportunity to explore many types of target-ligand interactions at the atomic level of detail based on complex crystal structures. Accordingly, exploring the wealth of currently available complex crystal structures and interaction information, a new genre of ...

Encoding Protein–Ligand Interaction Patterns in Fingerprints and Graphs

“…Among the most useful simplification processes for analyzing protein−ligand interactions is the conversion of atomic coordinates into simpler 1D or 2D fingerprints. 1 Fingerprints are easy to generate, manipulate, compare and, therefore, enable a systematic analysis of large data sets. They are largely used to describe and compare molecular objects (small molecular weight ligands, 2 pharmacophores, 3 proteins, 4 and protein−ligand binding sites 5 ) and represent descriptors utilized by computer-aided drug design programs, notably in silico screening tools.…”

“…Only the shortest interaction (number 1) between a single ligand atom and many protein atoms is kept (interaction 2 is not conserved). Remaining hydrophobic IPAs (1,3,4) are then clustered to yield IPAs 3 and 5. (C) Case of aromatic interactions between protein and ligand aromatic atoms (yellow balls).…”

Molecular Interaction Fingerprints

“…Assuming that similar structures share similar bioactivity, fingerprints have been frequently used to screen compound archives for similarity to reference ligands and therefore identify novel bioactive compounds [7]. This observation led to the design of ligand fingerprints focusing on protein interacting atoms [9]. One reason for this discrepancy is that all ligand atoms are not in direct interaction with host proteins.…”

Protein Interface Pharmacophore Mapping Tools for Small Molecule Protein: Protein Interaction Inhibitor Discovery

“…1). Where a lot of structural information about an interaction exists, a pharmacophore model can be built by mapping the key interactions between the ligand and the protein [28]. This can be further refined by superimposing multiple ligands in order to discover the common features [29].…”