Using molecular dynamics simulation to enhance conservation analysis for cross species extrapolation of the...

The U.S. Environmental Protection Agency's web-based Sequence Alignment to Predict Across Species Susceptibility (SeqAPASS) tool was developed to evaluate protein conservation across species through sequence and structural alignments to gather lines of evidence for predicting chemical susceptibility. Although SeqAPASS can rapidly generate predictions of species susceptibility in terms of a "yes" or "no" output, there is a growing interest in deriving more quantitative metrics for enhancing these predictions. To do this, a bioinformatics workflow was developed that combined SeqAPASS results with molecular docking and molecular dynamics (MD) simulations. This workflow was developed using transthyretin (TTR) and a per- and polyfluoroalkyl substance, with an emphasis on perfluorooctanoic acid (PFOA) as it is known that PFOA binds to TTR in humans and other experimental animals. This workflow was applied to generate quantitative information as additional lines of evidence for the conservation of the PFOA-TTR interaction across species. The SeqAPASS analysis predicted hundreds of species as susceptible based on conservation of the PFOA-TTR interaction (Level 1: 952 species, Level 2: 976 species, Level 3: 750 species). Predicted TTR structures from a subset of the species predicted as susceptible by SeqAPASS were used in molecular docking and MD simulations. The simulations supported that Lysine-15 is a key residue for the PFOA-TTR interaction. Quantitatively there was no significant difference in the species tested regarding their predicted binding affinities or other metrics specific to the chemical-protein interactions. These results demonstrated that the interaction between TTR and PFOA is likely conserved across various vertebrate taxonomic groups. Overall, this work provides a template for how advanced bioinformatics tools like MD simulations can be applied within ecotoxicology for improving cross-species predictions of chemical susceptibility. Importantly, our efforts aim to demonstrate applicability of these computational methods for integration in next-generation risk assessments.