An Entropy-Based Approach to Model Selection with Application to Single-Cell Time-Stamped Snapshot Data

Area of Science:

• Biophysics
• Systems Biology
• Computational Biology

Background:

• Single-cell experiments generate time-stamped snapshot (TSS) data, revealing cell-to-cell variability in protein abundances.
• TSS data offer insights into protein dynamics and signaling kinetics, but tracking individual cells over time is not feasible.
• Existing model selection methods struggle with the numerical and approximate nature of TSS data distributions.

Purpose of the Study:

• To develop a robust method for selecting the most accurate mechanistic model from TSS data.
• To overcome limitations of traditional model selection techniques in the context of noisy, numerically defined biological data.
• To enhance the understanding of signaling kinetics by accurately identifying underlying biological mechanisms.

Main Methods:

• Developed a generalized method of moments (GMM) based approach for parameter estimation using TSS data.
• Introduced an entropy-based model selection framework integrating GMM parameter estimation and kernel density estimation.
• Employed a bootstrap procedure to calculate model selection probabilities for assessing relative model support.

Main Results:

• The proposed entropy-based approach successfully identified the correct "ground truth" model from competing mechanistic models using simulated TSS data.
• Demonstrated the efficacy of the GMM-based parameter estimation in handling TSS data.
• Showcased the utility of bootstrap-computed model selection probabilities for quantifying support for candidate models.

Conclusions:

• The novel entropy-based method provides a reliable solution for mechanistic model selection with TSS data.
• This approach advances the analysis of single-cell dynamics and the elucidation of biological signaling pathways.
• The framework offers a valuable tool for researchers studying complex biological systems using high-throughput single-cell measurements.