A simple representation of three-dimensional molecular structure

Abstract

Statistical and machine learning approaches predict drug-to-target relationships from 2D smallmolecule topology patterns. One might expect 3D information to improve these calculations.Here we apply the logic of the Extended Connectivity FingerPrint (ECFP) to develop a rapid, alignment-invariant 3D representation of molecular conformers, the Extended ThreeDimensional FingerPrint (E3FP). By integrating E3FP with the Similarity Ensemble Approach (SEA), we achieve higher precision-recall performance relative to SEA with ECFP on ChEMBL20, and equivalent receiver operating characteristic performance. We identify classes of molecules for which E3FP is a better predictor of similarity in bioactivity than is ECFP.Finally, we report novel drug-to-target binding predictions inaccessible by 2D fingerprints and confirm three of them experimentally with ligand efficiencies from 0.442 -0.637 kcal/mol/heavy atom.

Filter feature selection for unsupervised clustering of designer drugs using DFT simulated IR spectra data

“…An Extended Three-Dimensional FingerPrint (E3FP) is motivated by ECFP that draws concentrically larger shells and encodes the 3D atom neighborhood patterns from small to larger shells iteratively. 37 Lastly, the Maximum Common Substructure (MCS) similarity is calculated by identifying structural overlap by matching atomic elements and bond types. 38 Tanimoto values calculated using binary fingerprints will always have a value between 0 and 1, with 1 indicating identical and 0 indicating entirely different.…”

Machine learning for target discovery in drug development

“…Given that SPiDER is built from two-ddimensional descriptors, one may expect a higher rate of false positive predictions for molecules with stereogenic centres relative to achiral molecules, as this information is not taken into account. To better predict the chiral nature of molecular recognition, the extended three-dimensional fingerprint was recently developed and applied to synthetic molecules [59], but remains to be validated with complex natural products.…”

Multi-reference poly-conformational computational methods for de-novo design, optimization, and repositioning of pharmaceutical compounds

“…The proposed computational algorithm extends the currently available methods [20][21][22][23] and introduces additional search flexibility via the use of the compound conformers. The proposal is to compare multiple possible shapes, adopted via varying environmental conditions, of the same molecule (i.e., conformers) rather than just a single shape that was used before.…”

“…An example of such an approach is ElectroShape implemented in the ODDT package [20] and is based on the algorithm that incorporates shape, chirality, and electrostatics [21,22], and represents each conformer via a fixed-length vector of real-valued numbers. Similarly the E3FP package [23] also utilizes an alignment-invariant 3D representation of molecular conformers as a fixed-length binary vector for each conformer. These fingerprint-based approaches allow to calculate the similarity between two molecular shapes either as a Tanimoto distance (for binary fingerprints) or Euclidean distance (for real-valued fingerprints) computations.…”