PreCisIon: PREdiction of CIS-regulatory elements improved by gene’s positION
Conventional approaches to predict transcriptional regulatory interactions usually rely on the definition of a shared motif sequence on the target genes of a transcription factor (TF). These efforts have been frustrated by the limited availability and accuracy of TF binding site motifs, usually represented as position-specific scoring matrices, which may match large numbers of sites and produce an unreliable list of target genes. To improve the prediction of binding sites, PreCisIon exploits the unrelated knowledge of the genome layout. Indeed, it has been shown that co-regulated genes tend to be either neighbors or periodically spaced along the whole chromosome. This novel type of information is combined with traditional sequence information by the PreCisIon machine learning algorithm. To optimize this combination, PreCisIon builds a strong gene target classifier by adaptively combining weak classifiers based on either local binding sequence or global gene position.
With the current state of the art, PreCisIon consistently improves methods based on sequence information only.
- Elati, M., Nicolle, R., Junier, I., Fernández, D., Fekih, R., Font, J., & Képès, F. (2012). PreCisIon: PREdiction of CIS-regulatory elements improved by gene’s positION. Nucleic acids research, gks1286.
- PreCisIon is part of the GREAT schema and can be found on the corresponding page on the GREAT web portal.