Designing novel building blocks is an overlooked strategy to improve compound quality.

One pragmatic way to improve compound quality, while enhancing and accelerating drug discovery projects, is the ability to access a high quality, novel, diverse building block collection. Here, we outline general principles that should be applied to ensure that a building block collection has the greatest impact on drug discovery projects, by discussing design principles for novel reagents and what types of reagents are popular with medicinal chemists in general. We initiated a program in 2009 to address this, which has already delivered three candidate drugs, and the success of that program provides evidence that focussing on building block design is a useful strategy for drug discovery.

Predictivity of Simulated ADME AutoQSAR Models over Time.

The automation of model building and model updating (autoQSAR) is an important step forward towards real-time small molecule drug discovery project support using the latest experimental data. We present here a simulation study using real company data of the behaviour of QSAR models over time. Three different global QSAR models, namely, human plasma protein binding, aqueous solubility and log D7.4 , are updated on a monthly basis over a period of three years. The effect of updating the models on their predictivity is studied using a series of monthly temporal test sets in addition to a final terminal temporal test set. Partial Least Squares (PLS), Random Forest (RF) and Bayesian Neural Networks (BNN) models are examined, covering three distinctly different approaches to QSAR modelling. It is demonstrated that the models are able to predict forward in time, but that updating models on a regular basis increases their ability to make predictions for current compounds. The degree of the improvement depends on the property studied and the model building technique used. These results demonstrate the importance of updating models on a regular basis. For both static models predicting forward in time, and regularly updating models it is shown that RF models are the most predictive for these data sets.

Identification of a putative intracellular allosteric antagonist binding-site in the CXC chemokine receptors 1 and 2.

The chemokine receptors CXCR1 and CXCR2 are G-protein-coupled receptors (GPCRs) implicated in mediating cellular functions associated with the inflammatory response. Potent CXCR2 receptor antagonists have been discovered, some of which have recently entered clinical development. The aim of this study was to identify key amino acid residue differences between CXCR1 and CXCR2 that influence the relative antagonism by two compounds that have markedly different chemical structures. By investigating the effects of domain switching and point mutations, we found that the second extracellular loop, which contained significant amino acid sequence diversity, was not important for compound antagonism. We were surprised to find that switching the intracellular C-terminal 60 amino acid domains of CXCR1 and CXCR2 caused an apparent reversal of antagonism at these two receptors. Further investigation showed that a single amino acid residue, lysine 320 in CXCR2 and asparagine 311 in CXCR1, plays a predominant role in describing the relative antagonism of the two compounds. Homology modeling studies based on the structure of bovine rhodopsin indicated a potential intracellular antagonist binding pocket involving lysine 320. We conclude that residue 320 in CXCR2 forms part of a potential allosteric binding pocket on the intracellular side of the receptor, a site that is distal to the orthosteric site commonly assumed to be the location of antagonist binding to GPCRs. The existence of a common intracellular allosteric binding site at GPCRs related to CXCR2 may be of value in the design of novel antagonists for therapeutic intervention.

QSAR modeling using automatically updating correction libraries: application to a human plasma protein binding model.

It is assumed that compounds occupying the same region of model space will be subject to similar errors in prediction, and hence, where these errors are known, they can be applied to predictions. Thus, any available measured data can be used to refine predictions of query compounds. This study describes the application of a correction library to a human plasma protein binding model. Compounds that have been measured since the model was built are entered into the library to improve predictions of current compounds. Time-series simulations were conducted to measure the time dependence of the correction library. This study demonstrates significant improvements in predictions where a library is applied, compared with both a static model and an updating model that includes recently measured data.