CaSPIAN: a causal compressive sensing algorithm for discovering directed interactions in gene networks.

Area of Science:

• Systems Biology
• Computational Biology
• Bioinformatics

Background:

• Inferring causal gene interactions is crucial for understanding cellular mechanisms.
• Existing methods often struggle with noisy biological data and require extensive prior information.

Purpose of the Study:

• To introduce CaSPIAN (Causal Subspace Pursuit for Inference and Analysis of Networks), a novel algorithm for accurate causal gene interaction inference.
• To evaluate CaSPIAN's performance on simulated and real biological networks, comparing it against existing algorithms.

Main Methods:

• Coupling compressive sensing with Granger causality techniques to identify sparse linear dependencies in gene expression time series.
• Utilizing a sequential subspace pursuit algorithm for reconstruction and Granger-type elimination for directionality estimation.
• Testing the algorithm with and without biological side-information (scaffold networks) from RNA-seq or microarray data.

Main Results:

• CaSPIAN demonstrated significant improvements in inference accuracy compared to related algorithms on simulated and real biological networks.
• Granger causality techniques effectively reduced false-positive interactions.
• Biological side-information enhanced sensitivity and precision, particularly in small sample scenarios.

Conclusions:

• CaSPIAN offers a computationally efficient and conceptually simple approach for inferring causal gene networks.
• The algorithm is robust to noisy measurements and benefits from biological priors for improved accuracy.
• This work highlights the utility of Granger causality in gene network inference.