Maximum likelihood estimation of log-affine models using detailed-balanced reaction networks

Area of Science:

• Biochemistry
• Computational Biology
• Systems Biology

Background:

• Determining the computational capabilities of biochemical systems is a key challenge.
• Log-affine models are used to represent biochemical reaction networks.
• Maximum Likelihood Estimation (MLE) is crucial for parameter estimation in these models.

Purpose of the Study:

• To extend existing constructions for generating mass-action systems with a unique positive steady state corresponding to the MLE.
• To investigate the impact of different spanning sets for the kernel of the design matrix on network dynamics.
• To identify conditions that guarantee the global stability of the MLE steady state.

Main Methods:

• Revisiting Gopalkrishnan's construction for mass-action systems.
• Extending the construction to utilize any finite spanning set of the kernel.
• Analyzing network properties such as boundary steady states, network deficiency, and convergence rates.
• Proving stability properties using specific spanning sets.

Main Results:

• The construction of mass-action systems with MLE as a unique positive steady state was generalized.
• The choice of spanning set significantly influences network dynamics, including steady-state behavior and convergence.
• Using a Markov basis as the spanning set guarantees the global stability of the MLE steady state.

Conclusions:

• The generalized construction provides a flexible framework for studying biochemical systems and parameter estimation.
• Markov bases offer a robust method for ensuring the stability of MLE steady states in log-affine models.
• This work advances the understanding of molecular computation and the analysis of complex biochemical networks.