Abstract: RNAs are versatile molecules that are involved in many important cellular activities including protein synthesis, antisense hybridization, RNA-RNA interactions, and RNA-protein interactions. Computational prediction of RNA higher-order structure is important because crystal structures have been determined only for a few RNA molecules. Statistical algorithms have been recently developed and shown to have advantages for RNA folding prediction. A statistical algorithm is presented for the energy model of base pair stacking, and a very important algorithm for more realistic energy rules is described. These algorithms demonstrate how statistical thinking can be successfully adopted for RNA secondary structure prediction to overcome inherent limitations in mathematical algorithms. For the determination of gene function, antisense techniques promise to offer a high-throughput platform. We illustrate how an approach based on statistical algorithms can be used for the rational design of antisense oligonuleotides (oligos). In the post-genomic era, DNA expression arrays and single-nucleotide polymorphisms (SNPs) promise to enable the prediction of gene functions and the identification of candidate genes for disease phenotypes. Functional predictions will eventually require experimental validation. An antisense approach is well suited for these high throughput applications to keep pace with rapid accumulation of genomic information, DNA expression array data, and SNP databases. The full realization of the promise of antisense technology can be greatly aided by an adequate integration of computational approaches and experimental techniques.
Key words and phrases: Antisense design, drug target validation, functional genomics, RNA folding.