Alternative Models

Alternative Model distinguishability algorithms & the software app DISTING

  • Date: January 3, 2017
  • People: Natalie R. Davidson, Keith R. Godfrey, Faisal Alquaddoomi, David Nola and Joseph DiStefano III
  • Primary Citations: Davidson et al.  Computer Programs in Biomedicine. In press.
  • Website: http://biocyb1.cs.ucla.edu/DISTING/

The model structural indistinguishability problem is often unfamiliar or ignored by modelers, which can lead to incorrect models and false predictions.  Algebraically, the problem is computationally complex, involving structured symbolic (non-numerical) analysis of typically large matrices. Until now, no software tools have been made available to address this problem. Our complex algorithmic solution for finding all  structurally identifiable alternative linear compartmental models that are input-output equivalent to a given structure is implemented in a new internet app DISTING online at this website.

The figure below illustrates DISTING output results for a particular 4-compartment model.   All models shown are input-output equivalent, meaning all would give identical outputs for the same inputs. 

The DISTING Web App

DISTING results for 4-compartment model.  All models  are input-output equivalent