Multistep Reaction Based De Novo Drug Design: Generating Synthetically Feasible Design Ideas.

We describe a "multistep reaction driven" evolutionary algorithm approach to de novo molecular design. Structures generated by the approach include a proposed synthesis path intended to aid the chemist in assessing the synthetic feasibility of the ideas that are generated. The methodology is independent of how the design ideas are scored, allowing multicriteria drug design to address multiple issues including activity at one or more pharmacological targets, selectivity, physical and ADME properties, and off target liabilities; the methods are compatible with common computer-aided drug discovery "scoring" methodologies such as 2D- and 3D-ligand similarity, docking, desirability functions based on physiochemical properties, and/or predictions from 2D/3D QSAR or machine learning models and combinations thereof to be used to guide design. We have performed experiments to assess the extent to which known drug space can be covered by our approach. Using a library of 88 generic reactions and a database of ∼20 000 reactants, we find that our methods can identify "close" analogs for ∼50% of the known small molecule drugs with molecular weight less than 300. To assess the quality of the in silico generated synthetic pathways, synthesis chemists were asked to rate the viability of synthesis pathways: both "real" and in silico generated. In silico reaction schemes generated by our methods were rated as very plausible with scores similar to known literature synthesis schemes.

NovoFLAP: A ligand-based de novo design approach for the generation of medicinally relevant ideas.

NovoFLAP is a computer-aided de novo design tool that generates medicinally relevant ideas for ligand-based projects. The approach combines an evolutionary algorithm (EA-Inventor) with a powerful ligand-based scoring function that uses both molecular shape and pharmacophore features in a multiconformational context (FLAP). We demonstrate that NovoFLAP can generate novel ideas that are not only appealing to design scientists but are also validated by comparison to compounds known to demonstrate activity at the desired biological target. NovoFLAP provides a novel computer-aided design technique that can be used to generate ideas that maintain desirable molecular attributes, such as activity at the primary biological target, while offering opportunities to surmount additional design challenges. Application to the design of the first nonbasic 5HT(1B) antagonist is presented.

Sharing chemical information without sharing chemical structure.

Studies to assess the risks of revealing chemical structures by sharing various chemical descriptor data are presented. Descriptors examined include "Lipinski-like" properties, 2D-BCUT descriptors, and a high-dimensional "fingerprint-like" descriptor (MACCs-vector). We demonstrate that unless sufficient precautions are taken, de novo design software such as EA-Inventor is able to derive a unique chemical structure or a set of closely related analogs from some commonly used descriptors. Based on the results of our studies, a set of guidelines or recommendations for safely sharing chemical information without revealing chemical structure is presented. A procedure for assessing the risk of revealing chemical structure when exchanging chemical descriptor information was also developed. The procedure is generic and can be applied to any chemical descriptor or combination of descriptors and to any set of structures to enable a decision about whether the exchange of information can be done without revealing the chemical structures.

Virtual screening for R-groups, including predicted pIC50 contributions, within large structural databases, using Topomer CoMFA.

Multiple R-groups (monovalent fragments) are implicitly accessible within most of the molecular structures that populate large structural databases. R-group searching would desirably consider pIC50 contribution forecasts as well as ligand similarities or docking scores. However, R-group searching, with or without pIC50 forecasts, is currently not practical. The most prevalent and reliable source of pIC50 predictions, existing 3D-QSAR approaches, is also difficult and somewhat subjective. Yet in 25 of 25 trials on data sets on which a field-based 3D-QSAR treatment had already succeeded, substitution of objective (canonically generated) topomer poses for the original structure-guided manual alignments produced acceptable 3D-QSAR models, on average having almost equivalent statistical quality to the published models, and with negligible effort. Their overall pIC50 prediction error is 0.805, calculated as the average over these 25 topomer CoMFA models in the standard deviations of pIC50 predictions, derived from the 1109 possible "leave-out-one-R-group" (LOORG) pIC50 contributions. (This novel LOORG protocol provides a more realistic and stringent test of prediction accuracy than the customary "leave-out-one-compound" LOO approach.) The associated average predictive r(2) of 0.495 indicates a pIC50 prediction accuracy roughly halfway between perfect and useless. To assess the ability of topomer-CoMFA based virtual screening to identify "highly active" R-groups, a Receiver Operating Curve (ROC) approach was adopted. Using, as the binary criterion for a "highly active" R-group, a predicted pIC50 greater than the top 25% of the observed pIC50 range, the ROC area averaged across the 25 topomer CoMFA models is 0.729. Conventionally interpreted, the odds that a "highly active" R-group will indeed confer such a high pIC50 are 0.729/(1-0.729) or almost 3 to 1. To confirm that virtual screening within large collections of realized structures would provide a useful quantity and variety of R-group suggestions, combining shape similarity with the "highly active" pIC50, the 50 searches provided by these 25 models were applied to 2.2 million structurally distinct R-group candidates among 2.0 million structures within a ZINC database, identifying an average of 5705 R-groups per search, with the highest predicted pIC50 combination averaging 1.6 log units greater than the highest reported pIC50s.

A knowledge-based approach to generating diverse but energetically representative ensembles of ligand conformers.

This paper describes a new and efficient stochastic conformational sampling method for generating a range of low-energy molecule conformations. Sampling can be tailored to a specific structural domain (e.g., peptides) by extracting torsional profiles from specific datasets and subsequently applying them to target molecules outside the reference set. The programs that handle creation of the knowledge-based torsional profiles and conformer generation per se are separate and so can be used independently or sequentially, depending on the task at hand. The conformational ensembles produced are contrasted with those generated using local minimization approaches. They are also quantitatively compared with a broader range of techniques in terms of speed and the ability to reproduce bound ligand conformations found in complexes with proteins.

Naphtho[2,1-b][1,5] and [1,2-f][1,4]oxazocines as selective NK1 antagonists.

Previously we reported on the synthesis and properties of a series of highly potent piperidinyl 2-subsituted-3-cyano-1-naphthamide NK1 antagonists that includes 3 and 4. Here we report our efforts to alleviate a troublesome atropisomeric property of those derivatives by introduction of a tethering bridge that, in addition, could be used to lock the resulting cyclic derivatives in a purported NK1 pharmacophore conformation. Using 3 as a starting point, the naphtho[2,1-b][1,5]oxazocine, 17, was found to contain the optimal ring tether size (8) for retaining NK1 activity, was more NK1 versus NK2 selective, and reduced the number of atropisomers from four to two. Cyclic derivatives 29 and 32, which exist as essentially single atropisomers in the purported pharmacophore conformation, were prepared in the closely related naphtho[1,2-f][1,4]oxazocine series as part of an effort to use mono methyl substitution of the tethering bridge as a conformation stabilizing factor. Both 29 and 32 were found to be less active as NK1 antagonists than the non-methylated parent 28 possibly due to methyl group destabilization of receptor interaction. We discuss the above findings in the context of a previously proposed NK1 pharmacophore model and present a further refinement of that model.