Filter inference: A scalable nonlinear mixed effects inference approach for snapshot time series data

Area of Science:

• Computational Biology
• Systems Biology
• Statistical Modeling

Background:

• Biological systems exhibit inherent variability, crucial for complex behaviors.
• Nonlinear mixed effects (NLME) modeling is used to study this variability.
• Traditional NLME inference is computationally intensive for large datasets, especially snapshot data.

Purpose of the Study:

• To develop a computationally efficient method for estimating NLME model parameters from snapshot measurements.
• To overcome the limitations of traditional NLME inference for large-scale biological data.
• To enable robust analysis of biological variability in high-throughput experiments.

Main Methods:

• Introduced filter inference, a novel approach for NLME parameter estimation.
• Utilized simulated individuals to define an approximate likelihood for model parameters.
• Employed state-of-the-art gradient-based Markov Chain Monte Carlo (MCMC) algorithms, including the No-U-Turn Sampler (NUTS).

Main Results:

• Filter inference provides an efficient alternative to traditional NLME methods for snapshot data.
• The method scales effectively with the number of model parameters.
• Demonstrated successful application in early cancer growth and epidermal growth factor signaling pathway modeling.

Conclusions:

• Filter inference makes NLME parameter estimation tractable for large snapshot datasets.
• This novel method enhances the analysis of biological variability in high-throughput biological studies.
• The approach offers a scalable and efficient solution for complex biological modeling challenges.