Dissolved organic matter (DOM) is one of the most important ligands governing the geochemical cycling of metals in the environment, but recent studies with Fourier transform ion cyclotron resonance mass spectrometry (FT-ICR-MS) have shown enormous complexity and diversity of DOM composition. How the diverse molecular composition of DOM affects the reactivity of DOM with metals is still largely unknown, which precludes us from developing accurate geochemical models for the fate of metals in the environment. In this study, we combined FT-ICR-MS analysis and theoretical modeling approaches and specifically elucidated the link between molecular composition and the proton and Cu binding ability of DOM, using the Suwannee River fulvic acid (FA) as a model humic substance. Batch adsorption experiments were conducted to generate different extents of molecular fractionation of FA samples by ferrihydrite. FT-ICR-MS analyses were employed to investigate the changes of molecular composition while Cu titration and the Windermere Humic Aqueous Model (WHAM) were used to quantify the variations on the Cu binding capacities of FA samples. We developed a general theoretical modeling approach, which integrated a suite of theoretical modeling methods, including the Vienna Soil-Organic-Matter-Modeler (VSOMM), SPARC Performs Automated Reasoning in Chemistry (SPARC), and the linear free energy relationships (LFER), for molecular modeling based on FT-ICR-MS data. Based on the FT-ICR-MS results, we found that, despite of the complex molecular composition of FA, FA molecules can be divided into three representative groups and each group of molecules had distinct chemical properties. Interestingly, molecules within the same group had similar distributions of molecular properties. Based on the chemical properties of the three groups of FA molecules, we successfully constructed three molecular models of FA using VSOMM, and quantified the distributions of proton and Cu binding constants with SPARC and LFER. Those independently determined binding constants were comparable to the WHAM default proton and Cu binding constants, supporting the validity of our modeling approach. Our modeling results suggested that the molecular complexity of DOM may be simplified with representative groups of molecules based on their binding ability with metals in theoretical modeling. Our modeling approach based on FT-ICR-MS data shed light on developing mechanistical models for metal reactions with DOM based on the molecular data, which is helpful for predicting the geochemical cycling of carbon and metals in the environment.