A Cheminformatics\nWorkflow to Select Representative\nTSCA Chemicals for New Approach Methodology (NAM) Screening

The Toxic Substances Control Act\n(TSCA) requires the US EPA to\nevaluate the hazard and exposure of new and existing chemicals. New\nchemical notifications are typically data-poor and EPA has historically\nrelied upon approaches including chemical categories to fill data\ngaps. As part of a multi-year Research Program, opportunities are\nbeing explored to leverage New Approach Methods (NAMs) in hazard and\nexposure assessments. Data from a battery of <i>in vitro</i> NAMs will be generated to form a case study for an adaptable approach\nto inform new chemical assessments. Herein, a cheminformatics workflow\nwas developed to identify a set of ∼300 representative candidate\nchemicals for <i>in vitro</i> screening from the TSCA non-confidential\nactive inventory. The freely available web application ClassyFire\nwas used to categorize all discrete organic structures from the TSCA\ninventory into one of 68 primary structural categories. Large primary\ncategories were subcategorized into smaller categories using hierarchical\nagglomerative clustering, ultimately yielding 180 structural terminal\ncategories. The inventory was filtered to substances that lacked previous\nToxCast bioactivity screening, were associated with physicochemical\nproperty predictions indicating non-volatile solids or liquids, and\nhad a higher chance of procurement. Amenability predictions for liquid\nchromatography–mass spectrometry were also generated to provide\nan indication of which chemicals lent themselves to aqueous-based\nscreening and analytical verification in solvated samples. Structures\nassociated with transformation in solvent, potentially explosive or\nhighly reactive, were excluded. Potential candidate substances were\nselected on the basis of being structurally representative of the\nterminal category and meeting other screenability conditions. A final\nset of 318 candidate chemicals were proposed to undergo analytical\nquality control and screening in a range of broad and targeted biological\ntechnologies for human health-relevant end points. Finally, <i>in silico</i> tools were applied to explore predicted hazard\nprofiles of these candidate substances relative to the full inventory.