The Stochastic System Identification Toolkit (SSIT) to model, fit, predict, and design experiments.

Area of Science:

• Computational Biology and Bioinformatics
• Systems Biology
• Biophysics

Background:

• Biological data exhibits intrinsic and extrinsic noise, leading to variability in experimental replicas.
• Stochasticity and heterogeneity in biological systems can contain valuable information but may bias parameter estimates if not properly modeled.
• Existing modeling approaches like lengthy simulations or deterministic averaging have limitations in accuracy and error tracking.

Purpose of the Study:

• To introduce the Stochastic System Identification Toolkit (SSIT), a software package for modeling, simulating, and solving chemical reaction models.
• To provide tools for parameter fitting, sensitivity analysis, and experimental design that account for biological noise and measurement errors.
• To enable efficient and accurate analysis of stochastic biological systems, improving prediction and experiment design.

Main Methods:

• The SSIT utilizes MATLAB's computational architecture for diverse modeling approaches, including ODEs, moments, Stochastic Simulation Algorithm (SSA), and Finite State Projection (FSP) of the Chemical Master Equation (CME).
• It incorporates probabilistic distortion operators to handle experimental noise and measurement errors.
• The toolkit offers advanced features like model reduction, joint fitting of models/datasets, sensitivity analysis, and Fisher information quantification.

Main Results:

• The SSIT provides a fast, flexible, and open-source solution for stochastic system modeling and analysis.
• Demonstrated application on yeast cell response to osmotic shock (mRNA counts) and breast cancer cell gene expression (single-cell RNA sequencing) highlights its utility.
• The software facilitates efficient parameter estimation, sensitivity analysis, and sequential experiment design.

Conclusions:

• The SSIT empowers researchers to build, simulate, and solve complex stochastic models with rigorous error tracking and efficient computation.
• It enables informed, time- and cost-effective experimental design by providing accurate parameter inference and predictive capabilities.
• The toolkit's graphical user interface and adaptable pipelines enhance accessibility for users across various scientific domains.