**What is CaSPIAN?**

Motivated by recent advances in compressive sensing theory and its application in practice, we introduce the concept of causal compressive sensing and design new **greedy list-reconstruction algorithms** for inference of causal gene interactions; as part of the process, we generate two sparse models for each potential interaction pattern and infer causality by comparing the residual errors of the models using statistical methods. Furthermore, in compressive sensing, the most difficult task consists of finding the support (i.e. the nonzero entries) of a sparse signal. This is accomplished by inferring the subspace in which the vector of observation lies. As a result, the complicated process of choosing the regularization coefficient in Lasso is substituted by the more natural task of choosing a “consistency” level between the vector of observations and its representation in the estimated subspace.

**Why use CaSPIAN?**

Causal compressive sensing can infer a relatively large fraction of causal gene interactions with very small false-positive rates when applied to small and moderate size networks. This finding is supported by simulated data, synthetic data from the IRMA network in Saccharomyces cerevisiae, and biological data from the human HeLa cell network and the SOS network of E. coli. The success probability is, as expected, highly influenced by the noise variance of the experiment and by the sampling time of the expressions. Our analysis of these phenomena adds to the understanding of the limitations of causal inference under imperfect measurement conditions, as well as the role of biological side information in reducing inference error rates. It also explains why available methods may not result in an improved detection probability upon adding as many time-shifted expression profiles as available, since gene expressions are usually measured at too widely separated times and have different time periods between the measurements.

**Where can I download CaSPIAN?**

You may access the CaSPIAN software from the following site here.

**Reference**

“CaSPIAN: A Causal Compressive Sensing Algorithm for Discovering Directed Interactions in Gene Networks”.

A. Emad and O. Milenkovic.

** Plos One**,

**, Open Access**

*31*(7)*(2014)*. [link]

**Contact**

Amin Emad (emad2@illinois.edu) and Olgica Milenkovic (milenkov@illinois.edu)