An Aggregation Advisor for Ligand Discovery

Colloidal aggregation of organic molecules is the dominant mechanism for artifactual inhibition of proteins, and controls against it are widely deployed. Notwithstanding an increasingly detailed understanding of this phenomenon, a method to reliably predict aggregation has remained elusive. Correspondingly, active molecules that act via aggregation continue to be found in early discovery campaigns and remain common in the literature. Over the past decade, over 12 thousand aggregating organic molecules have been identified, potentially enabling a precedent-based approach to match known aggregators with new molecules that may be expected to aggregate and lead to artifacts. We investigate an approach that uses lipophilicity, affinity, and similarity to known aggregators to advise on the likelihood that a candidate compound is an aggregator. In prospective experimental testing, five of seven new molecules with Tanimoto coefficients (Tc's) between 0.95 and 0.99 to known aggregators aggregated at relevant concentrations. Ten of 19 with Tc's between 0.94 and 0.90 and three of seven with Tc's between 0.89 and 0.85 also aggregated. Another three of the predicted compounds aggregated at higher concentrations. This method finds that 61 827 or 5.1% of the ligands acting in the 0.1 to 10 μM range in the medicinal chemistry literature are at least 85% similar to a known aggregator with these physical properties and may aggregate at relevant concentrations. Intriguingly, only 0.73% of all drug-like commercially available compounds resemble the known aggregators, suggesting that colloidal aggregators are enriched in the literature. As a percentage of the literature, aggregator-like compounds have increased 9-fold since 1995, partly reflecting the advent of high-throughput and virtual screens against molecular targets. Emerging from this study is an aggregator advisor database and tool ( http://advisor.bkslab.org ), free to the community, that may help distinguish between fruitful and artifactual screening hits acting by this mechanism.