Direct solution of the Chemical Master Equation using quantized tensor trains.

Area of Science:

• Computational Biology
• Biochemical Systems Analysis
• Numerical Mathematics

Background:

• The Chemical Master Equation (CME) is crucial for stochastic modeling of biochemical reactions but suffers from computational challenges.
• Existing methods to solve the CME often face the curse of dimensionality, limiting scalability with system size and complexity.

Purpose of the Study:

• To present a novel computational approach for solving the CME that overcomes the curse of dimensionality.
• To demonstrate the efficacy of Quantized Tensor Train (QTT) numerical methods for representing and solving the CME.

Main Methods:

• Utilized Quantized Tensor Train (QTT) numerical linear algebra for efficient tensor representation of the CME.
• Employed hp-discontinuous Galerkin discretization in time coupled with Density Matrix Renormalization Group (DMRG) algorithms.
• Developed an adaptive basis selection procedure to optimize computational complexity.

Main Results:

• The QTT-based method successfully represented the CME, reducing computational complexity.
• Demonstrated significant speedups and orders-of-magnitude storage savings compared to direct solution methods.
• Validated the approach on systems biology models including birth-death processes, futile cycles, and stochastic switches.

Conclusions:

• The proposed QTT method offers a powerful and scalable solution for the CME in systems biology.
• This approach effectively mitigates the curse of dimensionality, enabling the analysis of larger and more complex biochemical networks.
• The adaptive nature of the method ensures efficient and accurate capture of system dynamics.