RNAhybrid is a program that predicts multiple potential binding sites of miRNAs in large target RNAs. It identifies the energetically most favorable hybridization sites of a small RNA in a large RNA, while forbidding intramolecular base pairings. The algorithm has linear time complexity in the target length, allowing efficient searching of long targets. Statistical significance of predicted targets is assessed using extreme value statistics, Poisson approximation, and effective numbers of orthologous targets. RNAhybrid, along with RNAcalibrate and RNAeffective, is available for download and as a Web tool. The program was applied to predict Drosophila miRNA targets in 3'UTRs and coding sequences. It successfully identified previously predicted or experimentally verified targets and additional new ones. Results were analyzed statistically, and the program's ability to identify significant targets was demonstrated. RNAhybrid uses dynamic programming to calculate the minimum free energy (MFE) of hybridizations between miRNAs and target RNAs. It allows for the prediction of optimal and suboptimal, nonoverlapping binding sites. The program also allows for the imposition of constraints, such as requiring perfect helices in the 5'-end of the miRNA. The statistical significance of predicted targets is evaluated using extreme value distributions. Normalized MFEs are modeled as extreme value distributed, and p-values are assigned to these normalized MFEs. The effective number of orthologous sequences is used to combine Poisson p-values from orthologous targets, accounting for statistical dependencies between sequences. RNAhybrid was tested on large databases of potential target sequences, including 3'UTRs from Drosophila and Anopheles. It was able to identify significant targets with low E-values, indicating their potential for miRNA regulation. The program's ability to handle large datasets efficiently and its statistical methods make it a powerful tool for predicting miRNA targets. The program was also tested on coding sequences, where it identified potential binding sites with E-values up to 1.0. The results suggest that miRNAs may not regulate their targets through binding to coding sequences, as such binding sites are under-represented. Overall, RNAhybrid is a fast and effective tool for predicting miRNA/target duplexes, with statistical methods that allow for the assessment of the significance of predicted targets. It is available for download and as a Web tool, making it accessible for further research and analysis.RNAhybrid is a program that predicts multiple potential binding sites of miRNAs in large target RNAs. It identifies the energetically most favorable hybridization sites of a small RNA in a large RNA, while forbidding intramolecular base pairings. The algorithm has linear time complexity in the target length, allowing efficient searching of long targets. Statistical significance of predicted targets is assessed using extreme value statistics, Poisson approximation, and effective numbers of orthologous targets. RNAhybrid, along with RNAcalibrate and RNAeffective, is available for download and as a Web tool. The program was applied to predict Drosophila miRNA targets in 3'UTRs and coding sequences. It successfully identified previously predicted or experimentally verified targets and additional new ones. Results were analyzed statistically, and the program's ability to identify significant targets was demonstrated. RNAhybrid uses dynamic programming to calculate the minimum free energy (MFE) of hybridizations between miRNAs and target RNAs. It allows for the prediction of optimal and suboptimal, nonoverlapping binding sites. The program also allows for the imposition of constraints, such as requiring perfect helices in the 5'-end of the miRNA. The statistical significance of predicted targets is evaluated using extreme value distributions. Normalized MFEs are modeled as extreme value distributed, and p-values are assigned to these normalized MFEs. The effective number of orthologous sequences is used to combine Poisson p-values from orthologous targets, accounting for statistical dependencies between sequences. RNAhybrid was tested on large databases of potential target sequences, including 3'UTRs from Drosophila and Anopheles. It was able to identify significant targets with low E-values, indicating their potential for miRNA regulation. The program's ability to handle large datasets efficiently and its statistical methods make it a powerful tool for predicting miRNA targets. The program was also tested on coding sequences, where it identified potential binding sites with E-values up to 1.0. The results suggest that miRNAs may not regulate their targets through binding to coding sequences, as such binding sites are under-represented. Overall, RNAhybrid is a fast and effective tool for predicting miRNA/target duplexes, with statistical methods that allow for the assessment of the significance of predicted targets. It is available for download and as a Web tool, making it accessible for further research and analysis.