P versus Q: structural reaction coordinates capture protein folding on smooth landscapes

Minding your p's and q's has become as important to protein-folding theorists as it is for those being instructed in the rules of etiquette. To assess the quality of structural reaction coordinates in predicting the transition-state ensemble (TSE) of protein folding, we benchmarked the accuracy of four structural reaction coordinates against the kinetic measure P(fold), whose value of 0.50 defines the stochastic separatrix for a two-state folding mechanism. For two proteins that fold by a simple two-state mechanism, c-src SH3 and CI-2, the Phi-values of the TSEs predicted by native topology-based reaction coordinates (including Q, the fraction of native contacts) are almost identical to those of the TSE based on P(fold), with correlation coefficients of >0.90. For proteins with complex folding mechanisms that have especially broad, asymmetrical free energy barriers such as the designed 3-ankyrin repeating protein (3ANK) or proteins with distinct intermediates such as cyanovirin-N (CV-N), we show that the ensemble of structures with P(fold) = 0.50 generally does not include the chemically relevant transition states. This weakness of P(fold) limits its usefulness in protein folding studies. For such systems, elucidating the essential features of folding mechanisms requires using multiple reaction coordinates, although the number is still rather small. At the same time, for simple folding mechanisms, there is no indication of superiority for P(fold) over structurally chosen and thermodynamically relevant reaction coordinates that correctly measure the degree of nativeness.