Related Experiment Videos
Human microRNA prediction through a probabilistic co-learning model of sequence and structure
Jin-Wu Nam1, Ki-Roo Shin, Jinju Han
1Graduate Program in Bioinformatics, Seoul National University Seoul 151-744, Korea.
Nucleic Acids Research
|July 1, 2005
Summary
A new computational model, ProMiR, identifies novel microRNA (miRNA) genes by analyzing precursor structures and sequences. This approach significantly expands the known miRNA repertoire, revealing extensive regulatory networks in eukaryotic cells.
Area of Science:
- Genomics
- Molecular Biology
- Bioinformatics
Background:
- MicroRNAs (miRNAs) are crucial regulatory molecules, but current detection methods miss many low-abundance or divergent genes.
- Existing computational and experimental techniques have limitations in identifying the full spectrum of miRNA genes.
Purpose of the Study:
- To develop a novel computational model for identifying microRNA genes.
- To improve the sensitivity and specificity of microRNA gene prediction.
- To discover novel miRNA genes and assess their abundance in the human genome.
Main Methods:
- Introduced ProMiR, a probabilistic co-learning model integrating precursor structure and sequence information.
- Validated ProMiR using 5-fold cross-validation on 136 human datasets.
- Screened human chromosomes 16-19 for novel miRNA gene candidates.
Main Results:
- ProMiR achieved 73% sensitivity and 96% specificity in classification.
- Identified at least 23 novel miRNA gene candidates with low sequence similarity to known miRNAs.
- Experimentally confirmed 9 candidates processed by Drosha, indicating ~40% prediction accuracy.
Conclusions:
- The ProMiR model effectively identifies novel miRNA genes, including distantly homologous ones.
- The miRNA gene family is likely larger than previously estimated.
- These findings suggest extensive, previously uncharacterized regulatory networks governed by miRNAs in eukaryotes.