Related Experiment Videos
A computational approach to identify genes for functional RNAs in genomic sequences
R J Carter1, I Dubchak, S R Holbrook
1Computational and Theoretical Biology Department, Physical Biosciences Division, National Energy Research Scientific Computing Center, Lawrence Berkeley National Laboratory, 1 Cyclotron Road, Berkeley, CA 94720, USA.
Nucleic Acids Research
|September 28, 2001
Summary
Researchers developed a machine learning method to identify novel functional RNA (fRNA) genes in prokaryotic and archaeal genomes. This computational approach accurately predicts new RNA genes, aiding future experimental studies.
Area of Science:
- Genomics
- Bioinformatics
- Computational Biology
Background:
- Identifying novel functional RNA (fRNA) genes in genomic sequences computationally remains a significant challenge.
- Existing methods lack the success required for comprehensive genomic analysis.
Purpose of the Study:
- To develop and validate a machine learning approach for predicting novel RNA genes in unannotated prokaryotic and archaeal genomes.
- To improve the accuracy of RNA gene identification by integrating multiple prediction parameters.
Main Methods:
- Utilized neural networks and support vector machines to extract common features from known RNA sequences.
- Employed nucleotide composition, RNA sequence motifs, and free energy of folding as predictive parameters.
- Applied the developed method to diverse bacterial and archaeal genomes, including Escherichia coli.
Main Results:
- Achieved high prediction accuracy: 80-90% for bacteria and 90-99% for hyperthermophilic archaea.
- Demonstrated significant accuracy improvement by combining nucleotide composition with sequence motifs and folding energy.
- Identified several known fRNAs and hundreds of predicted novel RNA genes.
Conclusions:
- The developed machine learning approach offers a successful computational strategy for identifying novel RNA genes.
- Simple genomes harbor numerous unidentified RNAs that can be computationally predicted prior to experimental validation.
- Web-based access to predictions and a user interface are provided for public use.