• Citer

Study of Meta-analysis Strategies for Network Inference Using Information-Theoretic Approaches

par Pham, Ngoc Cam;Haibe-Kains, Benjamin ;Bellot, Pau;Bontempi, Gianluca ;Meyer, Patrick E.
Référence Proceedings - International Workshop on Database and Expert Systems Applications, page (76-83), 7816628
Publication Publié, 2017-01

Article révisé par les pairs

• Détails
• Contenu

Résumé :

Reverse engineering of gene regulatory networks (GRNs) from gene expression data is a classical challenge insystems biology. Thanks to high-throughput technologies, amassive amount of gene-expression data has been accumulatedin the public repositories. Modelling GRNs from multipleexperiments (also called integrative analysis) has, therefore, naturally become a standard procedure in modern computational biology. Indeed, such analysis is usually more robustthan the traditional approaches focused on individual datasets, which typically suffer from some experimental bias and a smallnumber of samples. To date, there are mainly two strategies for the problemof interest: the first one ('data merging') merges all datasetstogether and then infers a GRN whereas the other ('networksensemble') infers GRNs from every dataset separately and thenaggregates them using some ensemble rules (such as ranksumor weightsum). Unfortunately, a thorough comparison of thesetwo approaches is lacking. In this paper, we evaluate the performances of various metaanalysis approaches mentioned above with a systematic set ofexperiments based on in silico benchmarks. Furthermore, wepresent a new meta-analysis approach for inferring GRNs frommultiple studies. Our proposed approach, adapted to methodsbased on pairwise measures such as correlation or mutualinformation, consists of two steps: aggregating matrices of thepairwise measures from every dataset followed by extractingthe network from the meta-matrix.