Inferring consensus structure from nucleic acid sequences

This paper presents an unsupervised inference method for determiningthe higher-order structure from sequence data. The method isgeneral, but in this paper it is applied to nucleic acid sequencesin determining the secondary (2-D) and tertiary (3-D) structureof the macromolecule. The method evaluates position -positioninterdependence of the sequence using an information measureknown as expected mutual information. The expected mutual informationis calculated for each pair of positions and the chi-squaretest is used to screen statistically significant position pairs.In the calculation of expected mutual information, an unbiasedprobability estimator is used to overcome the problem associatedwith zero observation in conserved sites. A selection criterionbased on known structural constraints of the strongest interdependentposition pairs is applied yielding position pairs most indicativeof secondary and tertiary interactions. The method has beentested using tRNA and 5S rRNA sequences with very good results. Received on July 20, 1990; accepted on January 15, 1991