(575a) AI-Guided Redesign of Protein Pores to Facilitate Critical Materials from Electronic/ Battery Waste

Critical minerals are key for clean energy applications and essential for enabling a low carbon circular economy. Sustainable ion recovery and recycling platforms will enable lower-carbon mining and energy-efficient recycling of ions from wastes. Nature provides numerous examples of biomacromolecules capable of highly selective metal binding.

Recently, the discovery of lanmodulin protein (LanM) has offered high binding affinity and conditional kinetic lability to lanthanide ions. While small for a protein (12 kDa), LanM is much larger than chemical extractants, lowering the maximum theoretical adsorption capacity (per unit volume) of a lanmodulin-based extractants. Moreover, LanM wildtype protein is a non-specific binder to all La-atoms, thus offering little to no selectivity. To this end, we have identified specific helical protein pores which when optimally redesigned can be endowed with differential sensing and capture of intended rare earth elements (REE). As a start we have been able to use pore proteins for selective separation of Europium (Eu) from Ytterbium (Yb). We are deploying our AI-driven, large-language model-based, single sequence protein structure protein structure prediction platform (RGN2¹) for design of such versatile protein pores. Using a design-build-test-learn cycle we integrate and learn from both successful and failed mutants (tested experimentally for REE capture) to efficiently charter the sequence-function landscape of such protein pores where structure serves as a bio-aware encoder for the sequence.