Exploring the chemical space of pharmaceutical extractables and leachables

Extractables and leachables (E&Ls) are potential drug product impurities required to remain within safe exposure limits. This study compiles a dataset of over 1200 potential E&Ls and explores its chemical space using two-tier structural clustering, physicochemical profiling, and in silico toxicity predictions. The two-tier approach identifies 78 Tier-1 parent clusters that represent broad chemical classes; the Tier-2 stratification defines 323 subclusters, many of which are sparsely populated. Comparison with a proprietary dataset of reported E&Ls from FDA shows that the compiled dataset is highly representative of the E&L chemical space, with only few chemical classes absent (and each with only few chemicals). In silico profiling indicates that mutagenic compounds and strong/extreme skin-sensitizers represent only a small fraction of the compiled dataset, consistent with previous literature; moreover, the lack of systemic toxicity structural alerts in most substances suggests low predicted toxicity. The analysis also shows ∼50% of the substances in the compiled dataset have experimental toxicity data, which may be used for safety evaluations. Overall, this study analyzes the chemical diversity of E&Ls, highlights both high and sparsely populated chemical classes, and provides information for future read-across applications alongside insights to support evidence-based strategies for the safety assessment of E&Ls.