Software Supplement for "Large Scale Learning of Structure-Activity Relationships Using a Linear SVM and Problem-specific Metrics"

Please note that the fingerprinting tool listed in this section requires an SD file with valid 3D structures. A Java Runtime Environment Version 1.6 is required to run the programs. Don't forget to to add some additional memory, for example 512MByte (-Xmx512M). This value depends on the number of samples and dimensionality of the feature space.

Fingerprint tool for molecular graphs

It supports a variable hash space up to 231 patterns. The resulting sparse fingerprints are written in LIBSVM/LIBLINEAR format. Executable JAR, Requirements: JRE 1.6.

Download FingerPrinter
Download Tutorial data set
Download Tutorial data with cluster data

Modification of LIBLINEAR for vHTS

This machine is suitable for (multi-)classification using high-dimensional sparse feature vectors. The development is mainly driven by Lars Rosenbaum.  Executable JAR, Requirements: JRE 1.6.

The main features of this modification are

Download modified LIBLINEAR executable jar
Download Sources

Tutorial for Classification Workflow and External Prediction

Download Tutorial

Processing Tools Mentioned in Tutorials

Download SetSplitter.jar