Do higher-order moments improve inference of population dynamics?
- 1Max Planck Institute for Evolutionary Biology
Abstract
Fitting mathematical models of population dynamics to microbial time-series data allows us to estimate the ecological processes and interactions taking place in the microbiome. Repeated experiments of microbial systems yield replicates which slightly differ from each other. Some of this variability arises due to the fact that births and deaths occur at random. Most prior work focuses on fitting a deterministic mathematical model to the average across replicates. We use a stochastic model to fit the variability to the observed variability across replicates. Using a simulation-driven approach, we study the conditions under which our approach allows us to infer a larger fraction of ecological parameters correctly. We observe a substantial improvement in parameter inference. Lastly, our Bayesian approach not only allows us to incorporate prior information about the system, but also provides a distribution of parameters which conveys some idea of the uncertainty of the estimates.
Related
- Correcting for ascertainment bias in the inference of population structure
- Open-world fungal ITS embeddings improve higher-rank placement and cross-view retrieval, but percent identity is stronger on a matched ITS2 novelty benchmark
- Thinking too positive? Revisiting current methods of population genetic selection inference
- From Individual Trajectories to Population Densities: Weak-Form Inference of Heterogeneous Growth Laws
- abcneuralnet: Simulation-based Inference with Bayesian Neural Networks and Explainable AI for Population Genomics Inference
BibTeX
@misc{relekar2026higherorder,
title = {{Do higher-order moments improve inference of population dynamics?}},
author = {C. Relekar and A. de Azevedo-Lopes and M. Sieber and A. Traulsen},
howpublished = {bioRxiv},
year = {2026},
month = sep,
doi = {10.64898/2026.09.28.754680},
url = {https://www.biorxiv.org/content/10.64898/2026.09.28.754680v1},
note = {CC BY},
}
RIS
TY - UNPB
TI - Do higher-order moments improve inference of population dynamics?
AU - C. Relekar
AU - A. de Azevedo-Lopes
AU - M. Sieber
AU - A. Traulsen
PY - 2026
DA - 2026/09/28
PB - bioRxiv
DO - 10.64898/2026.09.28.754680
UR - https://www.biorxiv.org/content/10.64898/2026.09.28.754680v1
AB - Fitting mathematical models of population dynamics to microbial time-series data allows us to estimate the ecological processes and interactions taking place in the microbiome. Repeated experiments of microbial systems yield replicates which slightly differ from each other. Some of this variability arises due to the fact that births and deaths occur at random. Most prior work focuses on fitting a deterministic mathematical model to the average across replicates. We use a stochastic model to fit the variability to the observed variability across replicates. Using a simulation-driven approach, we study the conditions under which our approach allows us to infer a larger fraction of ecological parameters correctly. We observe a substantial improvement in parameter inference. Lastly, our Bayesian approach not only allows us to incorporate prior information about the system, but also provides a distribution of parameters which conveys some idea of the uncertainty of the estimates.
ER -