Protein structure-function relationship has been rethought over the past decades to account for conformational variation and its functional role. Coupled-folding and binding process in biomolecular recognition is a manifestation of such conformational disorder and heterogeneous conformational ensemble. The ensemble nature and the coupled dynamics nature between protein-water interactions, conformational fluctuations, and folding/binding events require high structural and temporal resolution, which create major experimental challenges. To address these challenges, this thesis presents a combined experimental and computational approach using two-dimensional infrared (2D IR) spectroscopy and computational spectroscopy. 2D IR spectroscopy offers sub-picosecond temporal resolution to probe protein structural variation. Site-specific structural information can be achieved by introducing isotope-labeling on selected amide groups, and computational spectroscopy that translates protein structures into an IR spectrum. To validate this approach, peptide-water interactions are studied on dialanine using 2D IR spectroscopy, and computational IR spectra predicted from molecular dynamics simulations. Amide I frequency fluctuation and vibrational energy relaxation is found to have the common origin of effective fluctuating forces due to water hydrogen bond dynamics. A chemical exchange process is also observed experimentally on the order of tens of ps, and is predicted to be coupled peptide-water motions through computational analysis. The idea of estimating conformational ensemble from isotope-labeled IR spectroscopy is tested on trialanine, which is well-characterized for its conformational variation. A Bayesian ensemble refinement scheme is developed for direct characterization of conformational ensemble against experimental IR spectroscopy. Isotope-edited 2D IR spectroscopy is found to provide a stringent constraint on the conformational distribution, and it returns consistent ensembles across different force fields and water models. The dominant factor influencing the quality of the ensemble refinements is the systematic frequency uncertainty from spectroscopic maps, but it can be significantly reduced by incorporating 2D IR spectra in addition to the Fourier-transform IR spectroscopy. This Bayesian ensemble refinement method with IR spectroscopy provide an effective approach to determine complex protein conformational ensemble. One of the model systems to understand coupled-folding and binding processes is association of human insulin monomer. During association, partially disordered B-chains from each monomer form an inter-monomer β-sheet in a native dimer crystal structure. However, the dimer conformational characterization still needs investigation. Conformational ensemble of insulin dimer is characterized using site-specific isotope-edited 2D IR spectroscopy, Markov State Models (MSMs), and amide I computational spectroscopy. Isotope-edited IR spectroscopy indicates an additional spectroscopic species other than the native dimer, and the distribution of these species can be influenced by tuning the ionic strength. The MSM predicted this additional conformation states of twisted dimer that exhibits a ~55° rotation of the native dimer interface, resulting in shifting the β-sheet registry and reorganizing its sidechain packing. Computational spectroscopy of the twisted dimer consistently accounts for the additional spectroscopic species. The presence of twisted conformations suggests a potential kinetic intermediate along the homodimer association and/or multi-pathway nature. This study provides additional insight on the conformational distribution of dimer and establishes a refined molecular picture of describing coupled folding and binding process in insulin monomer association.