A principle of stepwise maximization in the economy of the RNA folding process has been previously formulated as the principle of sequential minimization of conformational entropy loss (SMEL). This principle leads to a predictive folding algorithm rooted in an “adiabatic ansatz”. In this approximation, microstates are lumped up into base-pairing patterns (BPPs), each of which is treated as a quasi-equilibrium state, and folding pathways are resolved as series of BPP transitions. Within this ansatz, a marker for the expediency of folding is a coarse Shannon information entropy, σ, which has been shown to reach its absolute minimum within experimentally relevant time scales, well below the limit of thermodynamic times. A rigorous treatment developed in this work validates the adiabatic approximation. Thus, a Lagrangian is identified at a semiempirical microscopic level, which is the variational counterpart of the SMEL principle. The Lagrangian computation of σ is contrasted with the adiabatic computation, revealing the subordination of torsional microstate dynamics to BPP transitions within time scales relevant to folding. Furthermore, the expediency of the folding process is explained by showing that the time scale for saturation of the capacity to generate information within the coarse BPP description of conformation space is commensurate with biologically relevant time scales and it stems from the inherent Lagrangian structure of the dynamics at the semiempirical level of description of chain torsions.
