Efficient exact inference for dynamical systems with noisy measurements using sequential approximate Bayesian computation

Abstract

Motivation Approximate Bayesian computation (ABC) is an increasingly popular method for likelihood-free parameter inference in systems biology and other fields of research, as it allows analyzing complex stochastic models. However, the introduced approximation error is often not clear. It has been shown that ABC actually gives exact inference under the implicit assumption of a measurement noise model. Noise being common in biological systems, it is intriguing to exploit this insight. But this is difficult in practice, as ABC is in general highly computationally demanding. Thus, the question we want to answer here is how to efficiently account for measurement noise in ABC. Results We illustrate exemplarily how ABC yields erroneous parameter estimates when neglecting measurement noise. Then, we discuss practical ways of correctly including the measurement noise in the analysis. We present an efficient adaptive sequential importance sampling-based algorithm applicable to various model types and noise models. We test and compare it on several models, including ordinary and stochastic differential equations, Markov jump processes and stochastically interacting agents, and noise models including normal, Laplace and Poisson noise. We conclude that the proposed algorithm could improve the accuracy of parameter estimates for a broad spectrum of applications. Availability and implementation The developed algorithms are made publicly available as part of the open-source python toolbox pyABC (https://github.com/icb-dcm/pyabc). Supplementary information Supplementary data are available at Bioinformatics online.

Inferring cell motility in complex environments with incomplete tracking data

“…To this end, we fitted our cellular Potts model to the motility statistics obtained from our simulated data using the automated computational inference pipeline FitMultiCell [36]. Model fitting is based on an approximate Bayesian computation approach (pyABC) [27, 28], in which frequent simulations of selected parameter combinations (=particles) and the subsequent comparison of the stochastic model outcomes to the observed summary statistics successively leads to determination of the individual parameters upon successful convergence of the algorithm ( Figure 3a and Materials and Methods ). However, applying the standard approach to the simulated data was insufficient to describe the observed motility dynamics ( Figure 3b,c , Figure S2 ) and to infer the model parameters defining cell motility that were used for simulation ( Figure 3d ).…”

“…While most studies relied on manual and qualitative adaptations of the simulation models to inform model parameters, nowadays advanced methods allow for automatic data-driven parameter inference for complex individual cell based models using different types of data [6, 25, 26]. For example, the FitMultiCell -pipeline based on the integration of Morpheus [17] with the approximate Bayesian computation method pyABC [27, 28] allows parameter inference and simulation of multicellular systems as e.g. based on data informed by live-cell microscopy [6].…”

The chemical Langevin equation for biochemical systems in dynamic environments

“…While there has been significant progress towards computationally efficient methods to simulate systems with time-varying propensities, there still remains scope for improvements. Improvements are particularly important in the context of inference problems, which are typically associated with enormous computational burden, despite algorithmic advances in the area of inference [56][57][58][59] .…”

Anin vitro- agent-based modelling approach to optimisation of culture medium for generating muscle cells

“…The parameters γ, t M T f use , k lat and k nuc form function subset L 2 (α, β). These parameters were derived via statistical inference using an approximate Bayesian computation sequential Monte Carlo (ABC-SMC) method [23]. ABC approximates a posterior distribution for each parameter by comparing the distance between summary statistics provided by the ABM with those measured from experimental images using a distance function (δ(x)).…”