Cheminformatics has furnished a number of visualization methods for structure-activity datasets such as scatterplots of properties, clustering by related structural features, and tables of R-group fragments. Data is often available for thousands to millions of compounds, and so the molecules are generally treated as the primary index, with one or several activity/property measurements being analyzed for any given purpose. The growing availability of public data that has both quality and quantity (e.g. ChEMBL and selected portions of PubChem) measured against a large number of different targets is driving the need for new methods of visualization: it is increasibly practical to create a matrix of compound vs. assays/targets, where both dimensions are large.<br/><br/>This requires new supporting technologies for assay analysis, e.g. semantic annotation of bioassay protocols, in order to be able to sort, filter and cluster assays using the same kinds of algorithms as are routinely applied to chemical structures. Plotting compounds vs. assays can reveal interesting insights into polypharmacology, and we will present our most recent progress toward making this data accessible with a web-based interface. Furthermore, we can also delve into the task of clustering assay protocols by target and appropriate constraints and create a back-end process for building structure-activity models for each group. Model building is usually done for just a handful of datasets at a time, but we have the option of creating thousands of them, which introduces new opportunities and challenges for visualization. A single proposed compound can be represented with a ranked list of predictions, for targets and off-targets, which encourages the discovery of serendipitous effects, and appropriate warnings for likely problems. Larger scale analysis of many compounds vs. many model predictions also benefits from advances in visualization software, and methods for clustering and merging, in order to tease out as much information as possible about the structure-activity relationships for many different biological systems concurrently.<br/>

Expanding the target dimension: How to visualize a lot of models

253rd National Meeting (2017)

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