The prevalence of nonspecific or 'promiscuous' inhibitors that seem to be hits in multiple high-throughput screening (HTS) campaigns, but which turn out to be dead ends when attempts are made to optimize their activity, is a key problem in the field of HTS. Shoichet and colleagues, writing in Nature Chemical Biology, now describe two high-throughput assays that aim to address this issue by aiding the detection of promiscuous inhibitors.

Various explanations have been put forward to account for promiscuous compounds, including chemical reactivity and interference with assay readouts. Recent work from the Shoichet lab has also identified another possible mechanism to explain promiscuous inhibition: formation of colloid-like aggregates of the compounds, which sequester and thereby inhibit enzymes nonspecifically. Hits from HTS, leads and even some drugs seem to inhibit various enzymes through this mechanism at the micromolar concentrations typically used in HTS.

To facilitate investigation of the extent of this problem, the authors developed two rapid assays based on a standard 96-well format for detecting aggregate-based inhibition. The first assay exploits the detergent-sensitive nature of aggregate formation; compounds that only inhibit β-lactamase in the absence of detergent are considered likely to be promiscuous. The second assay uses dynamic light scattering to detect aggregate formation.

Shoichet et al. then selected 1,030 drug-like compounds and screened these molecules at micromolar concentrations in both assays, and a number of significant results were obtained. First, 19% of a subset of the compounds selected at random were detergent-sensitive inhibitors at screening-relevant concentrations (30 μm) — a percentage sufficiently high that it could dominate a screen that did not control for this effect. Second, both assays (whose reliability was confirmed using more sensitive, low-throughput versions of each assay) were able to robustly detect promiscuous aggregates, although of the two the detergent-sensitive assay seems best-suited for larger-scale applications. Finally, computational models for predicting aggregation-based promiscuity exploiting the results from the assays also showed some potential, and the data provided freely by the authors should aid the development of further such models.