DIGS is a pathway-level computational tool for disease classification using microarray gene expression profile. For each pathway specific gene expression profile, DIGS derives a new composite feature, called pathway activity, which summarises the expression patterns of its constituent genes. Pathway activity is constructed as a weighted linear summation of the expression profiles of its constituent genes, with the gene weights being optimised by DIGS so that the resulting pathway activity can optimally distinguish samples from different phenotypes. In DIGS, the maximum number of genes having non-zero weights can be controlled explicitly by user. The resulting activity vectors from all pathways are then assembled to form a pathway activity profile, on where a classifier can be trained to predict phenotype of new samples. DIGS is applicable to both two-phenotype and multiphenotype disease classification problems.

The following journal publication, entitled ”Pathway Activity Inference for Multiclass Disease Classification through Mathematical Programming Optimisation Framework”, by Lingjian Yang, Chrysanthi Ainali, Sophia Tsoka and Lazaros G. Papageorgiou, offers more details on the methodology of DIGS.

In order to run DIGS, both third party softwares GAMS(www.gams.com) and TORQUE (http://www.adaptivecomputing.com/products/open-source/torque/), together with their licences are required. DIGS makes calls to GAMS, which solves the underlying DIGS optimisation model to optimise the gene weights for each pathway. One of the mixed integer linear programming solvers, either CPLEX or GUROBI, must be available. TORQUE is used to schedule the batch jobs for solving multiple pathways. DIGS should be ran in Linux command line.

The following versions of DIGS are available: