Data-driven reconstruction of biological networks is a crucial step towards making sense of large volumes of biological data. While several methods have been developed recently for reconstruction of the networks, no comprehensive study has been carried out to compare these characteristically different methods in terms of their performance with regard to important aspects such as incomplete data-sets and noisy data. In this paper we have applied and compared four methods, viz. least squares (LS), principal component regression (PCR), linear matrix inequalities (LMI), and Least Absolute Shrinkage and Selection Operator (LASSO), on a real data set and a synthetic data set with respect to important metrics. This comparison gives us an insight into when to choose an appropriate approach for reconstruction of networks based on a priori properties of experimental data.

Contributed by Shankar Subramaniam.