Improving multi-scale characterization of nucleic acids and DNA nanotechnology toward therapeutic design - PROJECT SUMMARY Although generative artificial intelligence has resulted in seemingly infinite ideas for biomolecular design, the major bottleneck for translation of these ideas into the therapeutic realm remains the validation, characterization, and testing phases. Accurate and efficient prediction of the electronic structure and properties of biologically relevant macromolecular systems, such as nucleic acids, could lead to breakthroughs in biophysical characterization. Unfortunately, due to the extensive size of these biological macromolecules, ab initio quantum chemistry methods are far out of reach. My research team at UNM has pioneered a combination of quantum chemistry (QC), molecular dynamics (MD), and machine learning (ML) that will enable the accurate prediction of electron densities, energies, and forces of biologically relevant macromolecular systems, including solvated nucleic acid molecules, and extended to biomolecular ligand-receptor complexes. We use a Euclidean neural network model to predict these electron densities, enabled by adequate configurational sampling using MD, extensive ab initio QC calculations of smaller components of these macromolecular systems, pieced together to form the ML training set. We propose to train a complementary ML model to accurately predict atomic energies and forces of these biomolecular systems, providing a full picture of the electronic-scale biophysical properties of these systems. These predicted densities, energies, and forces for solvated nucleic acid complexes will be used to study the binding energetics of ligand-receptor complexes at the ab initio level of theory, calculations which are typically performed using classical approximations even though the binding domain is not well-represented with these approximations. Additionally, my research team is in the process of expanding the design space for DNA nanotechnology to include higher-order junctions which have an extensive conformational landscape that could enable precise control over multi-dimensional nanoscale lattice parameters, and heterochiral DNA junctions (including both right- and left-handed DNA) which are resistant to in vivo degradation. While DNA nanotechnology is an excellent platform for therapeutic design due to its inherent non-toxicity and its ability to self-assemble in various shapes due to self-complementarity of base-pairing, the standard building block called the four-way junction is limited in conformation and sequence possibilities. In addition, typical right- handed DNA degrades quickly in vivo, limiting the application space for therapeutics. The expansion of the DNA nanotechnology design space toward precise, controllable, and adaptable lattices at the nanoscale that can resist in vivo degradation could enable a variety of important applications in biomedicine and therapeutics.
