Exposing hidden high-affinity states

An invisible state doing the binding

HIV-1 TAR is a small, flexible RNA that interconverts among conformations. The one that binds ligands most tightly is barely populated in the free RNA — so an equilibrium binding measurement reports an affinity that's an average over the ensemble, badly underestimating what the high-affinity state can actually do.

To target TAR rationally, you need the affinity and population of a state you can't directly see.

Let the data constrain the hidden states

We cast TAR as a six-state thermodynamic model and fit it to the NMR observables within a Bayesian framework — inferring not just point estimates but full posterior distributions over each state's population and binding affinity. The Bayesian treatment is essential here: it propagates the uncertainty from a sparsely-populated state honestly, rather than reporting an overconfident single number.

The result is a credible-interval picture of states the spectra only touch indirectly.

Nanomolar binding behind a conformational penalty

The analysis unmasked a high-affinity conformation with nanomolar binding that is almost absent in the free RNA — its apparent weakness is a conformational penalty, the free-energy cost of reaching the binding-competent state, not poor intrinsic affinity. Separating those two contributions reframes what a TAR-targeting ligand actually has to overcome, and where the design leverage really is.

Population of all six states vs. total ligand [L_T]. The high-affinity BLα state (cyan) stays below the ~5% detection limit (dashed) in free TAR and only peaks transiently; at saturation BL₂ (blue) and CL₂ (red) dominate.

The affinity was always there — it was the population that hid it. Tell those apart and the design problem changes.