A central issue in molecular biology is understanding the regulatory mechanisms that control gene expression. The availability of whole genome sequences opens the way for computational methods to search for the key elements in transcription regulation. These include methods for discovering the binding sites of DNA-binding proteins, such as transcription factors. A common representation of transcription factor binding sites is a position specific score matrix (PSSM). We developed a probabilistic approach for searching for putative binding sites. Given a promoter sequence and a PSSM, we scan the promoter and find the position with the maximal score. Then we calculate the probability to get such a maximal score or higher on a random promoter. This is the p-value of the putative binding site. In this way, we searched for putative binding sites in the upstream sequences of Saccharomyces cerevisiae, where some binding sites are known (according to the Saccharomyces cerevisiae Promoters Database, SCPD). Our method produces either exact p-values, or a better estimate for them than other methods, and this improves the results of the search. For each gene we found its statistically significant putative binding sites. We measured the rates of true positives, by a comparison to the known binding sites, and also compared our results to these of MatInspector, a commercially available software that looks for putative binding sites in DNA sequences according to PSSMs. Our results were significantly better. In contrast with us, MatInspector doesn't calculate the exact statistical significance of its results.
- Binding sites
- Transcription factors