Recent efforts have delivered selective probe molecules for bromodomain proteins to elucidate their biological roles in oncological, cardiovascular and immuno-inflammatory disorders at the individual protein level. Despite this, the design of selective bromodomain inhibitors remains a challenge. We present here the first proteochemometric (PCM) multi-target computational model to predict the bioactivity of small molecules for the bromodomain target family, with the aim to better understand selectivity. Unlike traditional single-target QSAR models, PCM uses both compound and protein information to predict activity for multiple related biological targets, inherently modelling similarity between targets, as well as the similarity between compounds.<br/>15,350 bioactivity data points across 31 bromodomains were integrated from AstraZeneca and public data, from which highly predictive random forest classification models were built, achieving an overall balanced accuracy on the test set of 0.91. We benchmarked the PCM models against alternative models and found that PCM outperforms global QSAR techniques (balanced accuracy 0.91 vs 0.72) and average per-target QSAR models (average balanced accuracy 0.83 vs 0.79), although local or global model preference varied according to the individual bromodomain being modelled.<br/>The model was used to identify compound sub-structures and binding site amino acid properties which drive activity for different bromodomains, including “binding site hotspots”, useful for future compound design. 1,139 compound-target pairs were experimentally validated across four bromodomains, namely BRD1, BRD4 BD1, BRD9 and BRPF1b, with performance evaluation for three subsets, including Interpolation Compounds, Selectivity Profiles and Singular Actives (balanced accuracies 0.66, 0.64, 0.29 respectively), identifying selective compounds, as well as 319 new actives hits, which can be predicted by the PCM method.

Computational selectivity modelling for bromodomains: Insights into selectivity and discovery of new small-molecule hits

256th National Meeting (2018)

The official archive and repository for all ACS Meetings & Expositions abstracts and select posters or presentations.