Local Energy Decomposition of Intramolecular Interactions: The CovaLED Approach and Its Application to Molecular Recognition in Biomolecular Assemblies.

Achieving quantitative insight into the noncovalent interactions that govern molecular function and biological activity in complex assemblies remains a major challenge for quantum chemical analysis, particularly in systems such as nucleic acids where standard force-fields are known to struggle. Local energy decomposition (LED) provides gold-standard coupled-cluster descriptions of intermolecular interactions but can become difficult to interpret chemically when interacting fragments are covalently connected, such as ligands embedded in nucleic acid frameworks or functional groups linked through bonding networks. In these situations, shared electron density across fragment boundaries obscures the physical interpretation of energy contributions. Here we introduce CovaLED, an extension of the LED scheme that enables a rigorous treatment of covalent connectivity within the LED framework. Application to nucleic-acid-based recognition systems demonstrates the capabilities of the approach. In a riboswitch RNA-ligand complex, CovaLED reveals how methylation of guanine leads to a loss of stabilizing hydrogen-bonding interactions, consistent with experimental binding affinity trends. In a segment of human DNA, the method enables accurate quantification of interactions between nucleotides covalently linked within the backbone. CovaLED thus enables coupled-cluster-level energy decomposition in realistic biomolecular systems, where covalent structure and noncovalent recognition are intrinsically intertwined.