--------------------
- Structural sensitivity of biological models revisited (vol 283, pg 83, 2011) doi link

Auteur(s): Cordoleani F.(Corresp.), Nerini D., Gauduchon M., Morozov Andrew, Poggiale J.-C.

(Article) Publié: Journal Of Theoretical Biology, vol. 300 p.376-376 (2012)


DOI: 10.1016/j.jtbi.2011.12.025
Résumé:

Enhancingthepredictivepowerofmodelsinbiologyisachallengingissue.Amongthemajor difficultiesimpedingmodeldevelopmentandimplementationarethesensitivityofoutcomesto variationsinmodelparameters,theproblemofchoosingofparticularexpressionsfortheparametriza- tion offunctionalrelations,anddifficultiesinvalidatingmodelsusinglaboratorydataand/orfield observations.Inthispaper,werevisitthephenomenonwhichisreferredtoasstructuralsensitivityofa model. Structuralsensitivityarisesasaresultoftheinterplaybetweensensitivityofmodeloutcomesto variationsinparametersandsensitivitytothechoiceofmodelfunctions,andthiscanbesomewhatofa bottleneckinimprovingthemodelspredictivepower.Weprovidearigorousdefinitionofstructural sensitivityandweshowhowwecanquantifythedegreeofsensitivityofamodelbasedonthe Hausdorffdistanceconcept.Weproposeasimplesemi-analyticaltestofstructuralsensitivityinanODE modelingframework.Furthermore,weemphasizetheimportanceofdirectlylinkingthevariabilityof field/experimentaldataandmodelpredictions,andwedemonstrateawayofassessingtherobustness of modelingpredictionswithrespecttodatasamplingvariability.Asaninsightfulillustrativeexample, we testoursensitivityanalysismethodsonachemostatpredator–preymodel,whereweuselaboratory data onthefeedingofprotozoatoparameterizethepredatorfunctionalresponse.