Reconstruction quality of a biological network when its constituting elements are partially observed

Victor Picheny (1) , Jimmy Vandel (1) , Matthieu Vignes (1) , Nathalie Villa-Vialaneix (1)

1 MIAT INRA - Unité de Mathématiques et Informatique Appliquées de Toulouse

Unravelling regulatory regulations between biological entities is of utmost importance to understand the functioning of living organisms. As the number of available samples is often very low (often less than one hundred), inference methods are frequently performed on a subset of variables which make sense in the mechanisms under study. Classical remedies are either data driven (e.g., differentially expressed genes) or knowledge driven (e.g., using ontology information). However, whatever the chosen solution, important variables are very likely missed by the selection process, which is the issue at stake in the present paper.

Domaines: Statistiques [math.ST]; Théorie [stat.TH]

Victor Picheny, Jimmy Vandel, Matthieu Vignes, Nathalie Villa-Vialaneix. Reconstruction quality of a biological network when its constituting elements are partially observed. Seventeenth International Conference on Artificial Intelligence and Statistics (AI & Statistics 2014), Apr 2014, Reykjavik, Iceland. pp.L014. ⟨hal-00982758⟩