Version 1.3 for eukaryotes
The new version now uses the new InterproScan 5 web interface.
Requirements
- TFpredict needs the latest Java™ JRE release (version 6 or later).
- TFpredict requires the tool BLAST+, which can be downloaded from ftp://ftp.ncbi.nlm.nih.gov/blast/executables/blast+/LATEST/
- TFpredict requires a command line interpreter (e.g., shell or DOS prompt), as no graphical user interface is provided.
Included third-party software
- We require the homology search tool BLAST+ developed by Camacho et al. as it is needed for generating our feature representations of protein sequences. Website: http://blast.ncbi.nlm.nih.gov/Blast.cgi?CMD=Web&PAGE_TYPE=BlastDocs&DOC_TYPE=Download
- We integrated the tool InterProScan developed by Zdobnov and Apweiler to detect protein domains in query protein sequences. Website: http://www.ebi.ac.uk/Tools/pfa/iprscan/
- We employed classifiers from the WEKA package developed by Hall et al. for TF/Non-TF discrimination and Superclass prediction. Website: http://www.cs.waikato.ac.nz/~ml/weka/index.html
Integrated data
We retrieved the labeled data needed for our supervised machine learning approach for TF/Non-TF classification and structural characterization from diverse databases. Specifically, TF sequences were obtained from the commercial, hand-curated databases TransFac (Biobase) and MATBASE (Genomatix). The two databases were unified, filtered for redundant entries, and their structural superclasses were obtained from TransFac. Furthermore, a dataset of Non-TF sequences was compiled by querying UniProt with specific keywords which are unambiguously referring to other functional classes of proteins (e.g., "histone", "kinase", "transmembrane protein", "chaperon", etc.). An overview of our data sources can be found here:
| Database | URL |
|---|---|
| TRANSFAC: | http://www.biobase-international.com/pages/index.php?id=transfac |
| MATBASE: | http://www.genomatix.de/online_help/help_matbase/matbase_help.html |
| UniProt: | http://www.uniprot.org |