Memorizing all significant foldings of a random RNA chain

Abstract

We implement a synchronous parallel neural network associating a two-state unit to each pair of bases in a random RNA chain in such a way that a Watson-Crick base-pair corresponds to an activated unit and the phase space flow realizes the search in the conformation space for the RNA chain. Accordingly, all significant secondary structures are associated to metastable or stable neural patterns and exploration of conformation space emerges as a collective property of a neural-like system made up of equivalent components. The synaptic tensor is constructed taking into account the cooperativity principles of secondary structure formation, incorporating nucleation effects for initiation of intra-chain duplex formation and stacking of base pairs. This tensor is readily reducible to Moore-Penrose normal form since al metastable patterns become stable and orthogonal in the T = 0 limit. The dynamics of the exploration of metastable patterns by the network are shown to follow a random energy model in accord with the already-established uncorrelated energy landscape for random RNA foldings.