Introduction

OSCAR (One-class SVM for Cis-elements Accurate Recognition) is a program that can be used to identify binding sites of known transcription factors on promoter regions. The algorithm is based on one-class support vector machine (One-class SVM). OSCAR uses the sequential compositoin of known binding sites, and further incorporates the locational preferences of binding events. For more details about OSCAR, please see the [manuscript of OSCAR] and [supplementary materials].

For questions about this web site or if you encounter any problems, do not hesitate to contact Bo Jiang . We are open to any suggestions or remarks concerning this web interface.

Input known Binding Sites

Input your konwn binding sites of a specific transcription factor in Fasta format (example, format). The location of a particular binding site with respect to some reference point (e.g. transcription start site) could be given if available, and this locational information can aid the recognition of potential binding sites. Note that the width of all input binding sites must be equal (if not, please fill in the gap with 'N') , and be specified in the refline.

Input Promoter Sequences

Promoter Region
Specify the promoter region to be screened (needed if the positional bias is considered):

from downstream to upstream .

DNA Sequences (Optional)
DNA Sequences should be submitted in Fasta format (example, format). Our program will screen both strands of promoter sequences to identify potential binding sites of the transcription factor. If the location of the first nucleotide with respect to a reference point is specified for each sequence in the refline, precise locations of potential binding sites will be given. Otherwise, the locations of predicted binding sites are given respect to the first nucleotide in each sequence.

Parameters

Sensitivity

A parameter v , which is ranging from 0 to 1, is used to control the sensitivity of the classifier The meaning of this parameter can be found in our paper.

Parameter v =

Number of Intervals

When positional bias is incorporated, the promoter region is split into several intervals. Please specify the number of intervals here. A large number allows for refined scanning, but is more susceptablt to over-fitting at the same time.

Number of intervals =