Conformational heterogeneity and vibrational signatures of dAMP in aqueous solution revealed by high-accuracy machine-learning potentials

Nucleotides are the fundamental building blocks of life, yet many aspects of their behavior and interactions in water remain poorly understood. Here, we investigate the structural and vibrational properties of 2^ 2 ′ -deoxyadenosine 5^ 5 ′ -monophosphate (dAMP) in aqueous solution using a machine-learning approach supported by IR spectroscopy experiments. We develop highly accurate neural network potentials through an active-learning protocol that systematically samples the free-energy landscape at the GGA level and efficiently transfers this information to higher levels of theory. This potential accurately describes dAMP in water over nanosecond timescales. By explicitly learning dipole moments, the simulations capture key spectral signatures while providing molecular-level insight into their origin. Analysis of the free-energy landscape and vibrational spectra indicates that dAMP in aqueous solution populates a mixture of conformations. These findings highlight the importance of conformational heterogeneity in nucleotides and establish a general framework for high-accuracy simulations of flexible biomolecules in solution.