header image left image
  Current Projects  
  Computer Facilities  
divider   Home   divider
divider   Imprint   divider   Sitemap   divider   Intranet   divider   Contact   divider   Links   divider
tug logo igb logo

ProClassify is a tool for proteomic data classification. It was intended originally for high-resolution mass-spectrometry data classification, but it can be of use for datasets of a completely different nature as well.

ProClassify utilizes well known learning algorithms for classification, but its most important feature is that before the start of the learning process it performs dimensionality reduction on the input data to select only those data elements which are important for the classification. This process, often called "feature selection", raises the classification accuracy on the one hand and hugely reduces the processing costs of the learning algorithm on the other. The dimensionality reduction algorithm that is currently used in ProClassify presumes that there are only two sample classes.

You can find several test datasets to use with ProClassify on the NCI Clinical Proteomics Program Databank page.

The MS data reduction algorithm used in this software is based on the MATLAB implementation by Stefano Ongarello.

This software includes the Daubechies D4 wavelet transform implementation developed by Ian Kaplan and the LibSVM library developed by Chih-Chung Chang and Chih-Jen Lin.

Thanks to Stefano Ongarello, Jochen Martin and Gerhard Thallinger for their help and support.

Developed by Ilya Boyandin 2005-2006
proclassify logo description tab