Title

Sensitivity analysis of Boolean networks

Topic

Conceive and evaluate the equivalent of the sensitivity analysis of dynamical systems but for Boolean models as used in Computational Biology

City and country

Palaiseau, France

Team or project in the lab

Lifeware team

Name and mail of the advisor

Sylvain Soliman Sylvain.Soliman@inria.fr

Name and mail of the head of the department

François Fages Francois.Fages@inria.fr

General presentation of the topic (roughly 5 to 10 lines)

Logical models (both Boolean and multi-level) are often used in Computational Biology when quantitative data is missing or when the scale of the model becomes so large that any quantitative analysis becomes unfeasible.

Those models, where the possible evolution of each species is defined by a logical formula, benefit from strong symbolic analyses, mostly based on SAT solving and Answer Set Programming. These allow to compute or approximate stable-states and even complex attractors even for very big models.

Objective of the internship (roughly 10 to 20 lines)

However, there is currently no general sensitivity analysis for such Logical models. So, even when a model's attractors correspond to experimental data, it is neither easy to know how robust the model is, nor what "parts" of it (logical formulae) are sensitive/controllable.

The aim of the internship is a preliminary study on Boolean models, where using stable-states as observable one defines and evaluates a distance between those of the "wild-type" original model, and those of a model with single Knock-Ins or Knock-Outs (i.e., a single formula is replaced by resp. true or false).

The "distance" used will be either a variant of a modified Hausdorff function or of the finite version of the Fréchet–Nikodým–Aronszajn pseudometric.

Some published logical models will be used for evaluation, and maybe even one that is currently developed by a Ph.D. student of one of our biologist collaborators.

Bibliographic references

Expected ability of the student

Some taste (rather than precise knowledge) about graphs, computational biology, symbolic approaches is necessary. The experiments will also imply writing some code in Python and/or Prolog, so some knowledge of those will come in handy.