Analysing biological networks of mixed-type data with copula models

Background and challenges

A particularly promising but as-yet unexplored approach for this purpose is the use of parametric copula models, which can be used to couple variables of disparate natures. The development of such a model in a computationally efficient graphical form thus represents an open methodological challenge for the inference of generic networks from mixed-type data.

Goals

The DINAMIC project aims to develop and implement an innovative and widely applicable multivariate framework based on copulas and random pairwise likelihood (Mazo et al., 2021) for the differential analysis of mixed-type networks.

These methodological developments will be based on a succession of three applications covering several research themes at INRAE:

• cognitive health networks in seniors following the introduction of nutritional supplements;
• phenotypic networks in response to thermal stress in maize lines structured according to their genetic proximity;
• multi-omic networks in sperm from groups of bulls with contrasting fertility.

Each application will motivate a distinct facet of our approach, highlighting the added value of our interdisciplinary collaboration. To combine theoretically sound and computationally efficient statistical developments with relevant modelling assumptions aligned with the underlying biology, the DINAMIC project relies on a continuous cycle of interactions between methodologists and domain-specific experts.

Our multivariate mixed-type network model will represent a new approach to digital biology, with the potential to generate new insights into network plasticity in a wide variety of scientific disciplines.

Contact - coordination :

Partnerships

INRAE participants

Partners