Thermodynamic properties of transfer RNA: a computational study

Abstract

By studying large samples of molecules on a statistical basis useful information can be obtained regarding the type of properties which are important for biological molecules, and the properties which have been selected by evolution. Here, computational algorithms for RNA secondary-structure prediction are used to analyse the thermodynamic properties of the complete set of transfer RNA sequences in the tRNA database. Significant differences are observed between the different classes of tRNA in the database, and these are strongly correlated with the content of C + G bases. For each class of tRNA, a set of random sequences is generated having the same base composition and the same length distribution as the real sequences. In each case the mean value of the minimum Gibbs energy for the random sequences is substantially higher than for tRNA sequences. Within the random sample, sequences with properties comparable to real tRNA molecules are rare. The probability that each ground-state base pair is present in thermal equilibrium is calculated. These probabilities are much higher for real sequences than random ones, indicating that real sequences have a much more stable secondary structure than random ones. The main reason for this is that there are fewer alternative structures with Gibbs energies close to the ground state in real sequences. Secondary structure in real tRNA is found to melt at higher temperatures than in random sequences, and over a narrower temperature range, i.e. the melting process is more cooperative. These results suggest that one important feature selected by evolution is the stability of the ground-state structure.