Negative Cooperativity in a Polybivalent Complex Allows the Prevalence of a Partial Bound State.
The central region of the cytoplasmic dynein complex, comprising the intermediate chain (IC) and two light chains (LC8 and Tctex1), has eluded thorough quantitative characterization due to its participation in a highly coupled seven-state binding network. Although isothermal titration calorimetry (ITC) is the gold standard for measuring binding thermodynamics, conventional analyses are limited to simple interaction schemes because individual isotherms contain insufficient information to resolve complex reaction networks. Here, we overcome this limitation by combining extensive experimental sampling with hierarchical Bayesian inference. We collected 39 ITC isotherms spanning eight experiment types and developed a global Bayesian framework integrating multiple datasets while explicitly accounting for concentration uncertainty. Using this approach, we fit the complete dataset to a mechanistic seven-state model, estimating 190 parameters, including 12 thermodynamic parameters while marginalizing over 178 nuisance parameters. Remarkably, this strategy yields 95% confidence intervals for thermodynamic values as narrow as 0.05 kcal/mol and back-propagates to nanomolar precision in effective concentrations, even when experimental concentrations are in the hundreds of micromolar. The resulting thermodynamic landscape enables predictive modelling of assembly populations under different scenarios, including binding states inaccessible to standard ITC analyses. These results reveal previously unrecognized binding states that may play key roles in dynein cargo attachment and release. More broadly, this work reveals a form of "multi-cooperativity" governing dynein assembly and demonstrates how intensive experimentation coupled with modern statistical tools can resolve complex molecular systems beyond the reach of traditional biophysical techniques. Significance StatementLarge, complex mechanistic processes have remained difficult to fully characterize, which limits interpretability of the underlying biology. We utilize a large dataset of 39 complementary experiments to fully characterize a seven-state system using Bayesian inference. This process achieves impressively precise fits with 0.05 kcal/mol width confidence intervals. The high precision enables assessment of simultaneous positive and negative cooperativity in the assembly of the dynein intermediate chain with its light-chain partners. Simulation of state populations suggests that this balancing cooperativity is finely tuned to allow access to a half-bound state which has been previously inaccessible quantitatively. Our approach is broadly applicable and supports an emerging principle of molecular regulation--negative cooperativity as a strategy for tuning responsiveness and dynamic control.