Knowledge-Inclusive Adaptive Physics-Informed Neural Network for Microbial Interaction Modelling

2026-06-04

AI Summary

This study addresses the limitations of conventional physics-informed neural networks (PINNs), which rely solely on experimental data for parameter inference and struggle to incorporate auxiliary knowledge such as textual information or network topology, thereby hindering the accuracy and interpretability of microbial interaction modeling. To overcome this, the authors propose an adaptive multi-source knowledge fusion framework built upon the generalized Lotka-Volterra model. This approach synergistically integrates metagenomic literature-derived text, microbial interaction network structures, and experimental abundance data to enable context-aware modeling of community dynamics and ecological interaction inference. By systematically introducing textual and graph-structured priors for the first time, the method achieves substantial performance gains: on both real and simulated human and plant microbiome datasets, it improves Bray–Curtis accuracy by 23%, R² by 47%, and overall performance by up to 53% compared to state-of-the-art methods.

Abstract

Physics-Informed Neural Network (PINN) is a way of including knowledge in the form of equations in Machine Learning methods. Beyond equations, knowledge exists in other forms, such as text and network structure. While existing PINN-based approaches discover equation parameters from data, they rely solely on experimental measurements. We propose a new PINN framework that enriches parameter discovery by incorporating auxiliary knowledge sources. We instantiate our framework for microbiology, where generalised Lotka-Volterra (gLV) serves as a biological foundation for modelling microbial communities. We demonstrate that incorporating knowledge improves microbial community modelling. Our framework enriches the gLV parameters using peer-reviewed metagenomics literature, as text provides biological context on external influences that gLV alone cannot capture. We combine this knowledge with experimental measurements of microbial abundance using a data-driven integration approach. We integrate network-based structural knowledge by explicitly modelling microbial interactions. Our knowledge-inclusive framework infers microbial networks, revealing ecological insights. We validate these findings against ecological roles documented in the literature. We evaluate on real and simulated datasets spanning human- and plant-associated microbial communities. Our framework improves over the state-of-the-art by up to 53%, even without knowledge. Knowledge addition yields gains of up to 23% in Bray-Curtis Dissimilarity-based accuracy and 47% in $\mathrm{R}^2$.

Problem

• Physics-Informed Neural Network
• Microbial Interaction
• Knowledge Integration
• Parameter Discovery
• Lotka-Volterra

Innovation

• Physics-Informed Neural Networks
• knowledge integration
• microbial interaction modeling
• generalized Lotka-Volterra
• multi-source knowledge fusion