Abstract
Isothermal titration calorimetry (ITC) is a powerful technique for probing biomolecular interactions. However, accurate determination of binding parameters-such as enthalpy and free energy-as well as associated uncertainties can be hindered by noise and concentration variability. Notably, the mathematical ambiguity surrounding analyte concentrations in standard binding models intrinsically limits the precision with which binding parameters, particularly binding enthalpies, can be determined. Here, we present a Bayesian pipeline that resolves this ambiguity by combining two key strategies: simultaneous analysis of multiple ITC datasets and a hierarchical Bayesian treatment of analyte concentration priors. This dual approach not only lifts the degeneracy inherent in single-dataset studies but also removes an ambiguity typically present in Bayesian analysis by self-consistently refining concentration estimates, ensuring optimal joint inference of binding parameters and concentrations. Using modern Monte Carlo techniques enables our pipeline to provide robust posterior sampling for more than 10 datasets and 40 total parameters. We validate the approach with synthetic ITC datasets for single- and multi-site binding models and apply it to experimental data, including 14 datasets for 1:1 binding of Mg(II) to the chelator EDTA and multiple datasets of the hub protein LC8 with diverse binding partners. This work serves as a foundation for improving the precision of binding constants using multiple ITC datasets, while providing a systematic framework for assessing the reliability of experimental concentration estimates, paving the way for more accurate biomolecular interaction studies.