(361i) Integrated Computational and Experimental Design of Selective Lanthanide Binding Peptides

Rare earth elements (REE), or lanthanides, are critical materials with a wide range of applications in clean energy industrysuch as battery, semiconductor, and electric vehicles, due to their unique luminescent, magnetic, and catalytic properties. Therefore, there is a growing demand for developing green industrial-scale extraction of lanthanides. However, efficient separation of heavy lanthanides (e.g. from Europium to Lutetium) with exceedingly similar chemical properties is technologically challenging, and is an active area of research. Here, we combine molecular dynamics (MD), enhanced sampling, and machine learning methods to design insilico short 17-residue peptides that selectively bind to trivalent REE cations in aqueous solution. The REE-peptide binding affinity are predicted by ML models that are trained on features that characterize the REE-peptide binding complex including hydration properties of REEs from literature studies, REE-peptide interaction energies obtained from MD simulations, and physicochemical properties of the peptide estimated from the DBAASP peptide database. Starting from the original lanthanide-binding tag (LBT) peptide, we computationally screen peptide candidates with mutations mainly on the six key ligating residues that improve predicted binding selectivity against 12 different REEs. Candidates with top selectivity scores selected from in silico discovery are verified by the fluorescence titration experiments. Moreover, we utilize a novel genetic algorithm framework to iteratively guide the design of new peptide sequences combining our model and experiments. Therefore, we demonstrate an integrated data-driven approach for quantitative sequence-structure-activity relationship as well as design of lanthanide binding peptides, paving the road for designing novel biomolecules for efficient rare earth elements recovery.