A key region of the dynein complex, involving the intermediate chain (IC) and two light chains (LC8 and Tctex) has eluded thorough characterization because together these three species comprise a seven-state binding model. Isothermal titration calorimetry (ITC) is the gold standard technique for binding thermodynamics but is typically limited to simple systems due to its inherent low information content. To overcome this limitation, we collected 39 different ITC isotherms across eight different experiment types and designed a hierarchical Bayesian inference pipeline designed for multiple data sets and concentration uncertainty to analyze these data. With this pipeline, we successfully fit the 39 isotherms to the seven-state reaction model, amounting to 190 parameters, including 178 nuisance and 12 thermodynamic, with 95% confidence intervals for thermodynamic values with widths as narrow as sub-percent. Strikingly, this extreme thermodynamic precision back-propagates, yielding nanomolar precisions for concentrations in the hundreds of micromolar. The extreme narrow precisions achieved for thermodynamic values enable confidence in modeling concentrations of the various products of the reaction for different scenarios, including binding states that were inaccessible by previous analyses. As a result, we hypothesize a functional relevance of these newly characterized binding states in IC binding and releasing cargo for the dynein complex. Our findings demonstrate the power of Bayesian inference in sharpening precision in low-information, yet high-diversity, experimental contexts.
Walker et al. (Sun,) studied this question.