LOCANDA: Exploiting causality in the reconstruction of gene regulatory networks

The reconstruction of gene regulatory networks via link prediction methods is receiving increasing attention due to the large availability of data, mainly produced by high throughput technologies. However, the reconstructed networks often suffer from a high amount of false positive links, which are actually the result of indirect regulation activities. Such false links are mainly due to the presence of common cause and common effect phenomena, which are typically present in gene regulatory networks. Existing methods for the identification of a transitive reduction of a network or for the removal of (possibly) redundant links suffer from limitations about the structure of the network or the nature/length of the indirect regulation, and often require additional pre-processing steps to handle specific peculiarities of the networks at hand (e.g., cycles). In this paper, we propose the method LOCANDA, which overcomes these limitations and is able to identify and exploit indirect relationships of arbitrary length to remove links considered as false positives. This is performed by identifying indirect paths in the network and by comparing their reliability with that of direct links. Experiments performed on networks of two organisms (E. coli and S. cerevisiae) show a higher accuracy in the reconstruction with respect to the considered competitors, as well as a higher robustness to the presence of noise in the data.