Synergizing Physically Constrained MCMC and Chemical-Informed Gaussian Processes for Reaction Network Discovery

Runzhe Liu, Zihao Wang, Wenbo Yang, Shengyang Tao · 2026

Read Synergizing Physically Constrained MCMC and Chemical-Informed Gaussian Processes for Reaction Network Discovery by Runzhe Liu, Zihao Wang, Wenbo Yang, Shengyang Tao (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2606-23757.txt.

Introduction

Liu, Wang, Yang, and Tao combine two Bayesian tools to recover chemical reaction networks from noisy time series data, in a 2026 paper on scientific machine learning. A reaction network states which reactions occur and how fast. Structure and rate constants are tightly coupled, so fitting one without the other invites physically impossible mechanisms that still match the data.

Their workflow, PC MCMC CIGP, samples candidate topologies with a spike and slab Markov chain method screened by hard conservation and thermodynamic rules. A chemical informed Gaussian process, which embeds the candidate mechanism as its prior mean, then calibrates parameters and guides which experiments to run next. On a hydrogen bromine benchmark the sampler separates true radical pathways from deceptive fits. On styrene epoxidation the optimization loop improves final yield by about twelve percent over a reported baseline.

Catalog record

CREATORRunzhe Liu, Zihao Wang, Wenbo Yang, Shengyang Tao
YEAR2026
COLLECTIONPAPERS
FORMATTXT
FILE SIZE44 KB
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2606.23757

Browse the full library · About the Tower of Babel

Full text

The opening of the work, from the library's hosted text:

Synergizing Physically Constrained MCMC and Chemical-Informed Gaussian Processes for Reaction Network Discovery Runzhe Liu, Zihao Wang, Wenbo Yang, Shengyang Tao Abstract Extracting interpretable governing equations from sparse, noisy chemical time-series data remains difficult because discrete reaction topology and continuous kinetic parameters are tightly coupled. We present PC-MCMC-CIGP, a reproducible gray-box workflow that combines spike-and-slab topology sampling, hard conservation and thermodynamic screening, and a Chemical-Informed Gaussian Process (CIGP) residual model for parameter calibration and experimental design. The methodological contribution is not a new MCMC or GP family in isolation; rather, it is the integration of these components into a physically constrained workflow with explicit uncertainty-aware acquisition choices. On the H 2 + Br 2 \mathrm{H_{2}+Br_{2}} benchmark, the constrained sampler distinguishes elementary radical pathways from deceptive phenomenological fits in our experiments. On styrene epoxidation, the CIGP optimization loop improves final yield by 12.5 % 12.5\% over the reported GP-BO baseline. A new 10-seed acquisition study shows that EI, GWU, PC-EI, uncertainty sampling, discrepancy hunting, and random search have different trade-offs: PC-EI substantially reduces low-yield BO suggestions, while EI-style criteria give the strongest final-yield performance. † † affiliationnotice: Equal contribution Code: github.com/Billy-Liu-DUT/pc-mcmc-cigp 1 Introduction The discovery of governing equations from noisy time-series data remains a fundamental challenge in scientific machine learning ( 4 ; 44 ; 7 ) . Although deep learning paradigms approximate continuous dynamics effectively, interpretability is often sacrificed for black-box interpolation ( 9 ) . In chemical kinetics, the primary objective is structure identification , defined as the reconstruction of the discrete reaction topology governing system evolution to recover a physically interpretable causal graph ( 6 ; 24 ) . Mechanism discovery is mathematically formulated as an ill-posed inverse problem characterized by the strong coupling between discrete structures and continuous kinetic constants ( 52 ; 13 ; 1 ) . Conventional Bayesian inference is frequently hindered by this interdependence. Standard parameter estimation techniques lack explicit physical encoding ( 54 ) , whereas structure sampling is often impeded by vast combinatorial spaces and insufficient physical constraints, leading to mathematically valid yet physically prohibited solutions ( 15 ) . To address these limitations, a unified physically constrained Bayesian framework is proposed to synergize discrete sampling on physical manifolds with chemically informed continuous modeling ( 3 ) . The problem is treated as the cooperative solution of structural inference and parameter estimation. A multi-physically constrained Spike-and-Slab Markov Chain Monte Carlo (MCMC) sampler is introduced for structural identification. Constraints regarding mass conservation, electron conservation, and thermodynamic potential loops are encoded as boundary conditions to prune the posterior distribution and eliminate spurious mechanisms. For parameter estimation, a Chemical-Informed Gaussian Process (CIGP) is utilized to construct a hybrid gray-box model. The mechanism-based ordinary differential equation is embedded as the prior mean function to capture dominant kinetic trends, while structural discrepancies are modeled via non-parametric kernels ( 41 ; 8 ; 32 ) . Physical parameters and kernel hyperparameters are jointly optimized to ensure robust estimation ( 10 ; 12 ) . Additionally, physics-aware acquisition functions are developed to facilitate automatic experimental design. Figure effig:framework summarizes the two-stage workflow: PC-MCMC first identifies a physically admissible reaction topology, and CIGP then uses the identified ODE model as a GP mean function during calibration and active learning. The main contributions of this work are summarized as follows: 1. A fixed-dimensional spike-and-slab MCMC workflow is formulated for candidate reaction selection, with mass, charge, and detailed-balance constraints used as hard admissibility checks. 2. A Chemical-Informed Gaussian Process calibration stage is used as a gray-box residual model, preserving the mechanistic ODE as the prior mean while modeling systematic discrepancy. 3. A reproducible active-learning benchmark is provided, including a 10-seed comparison of PC-EI, EI, GWU, discrepancy hunting, uncertainty sampling, and random search. Figure 1 : Overview of the proposed PC-MCMC-CIGP framework. The upper panel illustrates the physically constrained structure discovery stage, where candidate reaction networks are sampled using a Spike-and-Slab MCMC scheme subject to mass conservation and thermodynamic constraints. The lower panel depicts the Chemical-Informed Gaussian Process (CIGP), which embeds the discovered mechanistic ODE model as the GP mean function while modeling structural discrepancies non-parametrically. The unified framework enables simultaneous mechanism identification, parameter inference, and physics-aware active learning. 2 Related Work 2.1 Phenomenological versus Mechanistic Discovery Data-driven discovery of kinetic laws is commonly approached through symbolic regression or sparse identification techniques ( 4 ; 42 ; 44 ; 5 ) . Algorithms such as genetic programming and sparse regression can recover accurate macroscopic expressions from observational data, including fractional-order kinetics that arise in radical chain reactions ( 9 ; 35 ) . Their limitation for this work is that an accurate expression is not necessarily a reaction mechanism: a half-order rate law may fit \mathrm{HBr} formation while leaving the elementary radical topology and hidden intermediates unidentified. 2.2 Reaction-Network and Neural-ODE Models Reaction network reconstruction has also been formulated as deterministic superstructure selection using MINLP, flux-balance-style constraints, or hypergraph search ( 30 ; 45 ; 55 ; 37 ; 2 ) . These methods can encode stoichiometry explicitly, but they often provide limited posterior uncertainty over alternative topologies. Recent neural ODE and chemical-reaction neural-network models improve flexible dynamics learning for kinetic systems ( 27 ; 36 ; 8 ) . They are complementary to our objective: rather than learning an unconstrained neural vector field, PC-MCMC-CIGP assumes a candidate elementary-reaction set and estimates which reactions are active, so the output remains an interpretable subset of reaction steps. 2.3 Probabilistic Modeling and Experimental Design Gaussian Processes (GPs) are widely used as probabilistic surrogates for calibration and Bayesian optimization ( 54 ; 28 ; 48 ; 47 ) . Standard GP-BO is effective for yield maximization, but it usually treats the experiment as a black-box response surface. Physics-informed and hybrid GP models incorporate mechanistic structure into the mean or kernel ( 26 ; 41 ; 32 ; 10 ) . Our CIGP component follows this gray-box tradition. The novelty claimed here is therefore deliberately scoped: we provide a unified, physically constrained reaction-discovery and active-learning workflow, not a fundamentally new GP inference algorithm. 3 The PC-MCMC-CIGP Framework 3.1 Problem Formulation We consider a chemical system with N_{s} species and a pre-enumerated candidate set of N_{r} elementary reactions \mathcal{R}=\{R_{j}\}_{j=1}^{N_{r}} . This candidate set is assumed known; the unknowns are the binary activity vector \boldsymbol{\gamma}\in\{0,1\}^{N_{r}} , the active reaction graph \mathcal{G}(\boldsymbol{\gamma}) , and kinetic parameters \mathbf{k} . Let \mathbf{c}(t)\in\mathbb{R}_{\geq 0}^{N_{s}} denote concentrations and let \mathbf{S}(\mathcal{G})\in\mathbb{Z}^{N_{s}\times N_{r}} be the stoichiometric matrix whose j -th column is the net stoichiometric vector of reaction R_{j} and is masked by \gamma_{j} during simulation. The mass-action dynamics are \frac{d\mathbf{c}}{dt}=\mathbf{S}(\mathcal{G})\,\mathbf{r}\bigl(\mathbf{c}(t);\mathbf{k},\boldsymbol{\gamma}\bigr), (1) where r_{j}(\mathbf{c};\mathbf{k},\boldsymbol{\gamma})=\gamma_{j}k_{j}\prod_{i}c_{i}^{\nu_{ij}^{-}} for reactant stoichiometric orders \nu_{ij}^{-} . Reaction rates follow either Arrhenius parameterization k=A\exp(-E_{a}/RT) or the detailed-balance parameterization described below ( 21 ; 51 ) . The mechanism-discovery data are sparse noisy trajectories \mathcal{D}=\{(t_{m},\mathbf{y}_{m})\}_{m=1}^{M} , where \mathbf{y}_{m} contains observed components of \mathbf{c}(t_{m}) under a Gaussian observation model. The active-learning stage uses a distinct notation: \mathbf{u}\in\mathcal{U} denotes controllable experimental conditions such as initial concentrations, temperature, and residence time, and y(\mathbf{u}) denotes the scalar objective such as product yield. This separates state variables \mathbf{c}(t) from design variables \mathbf{u} . 3.2 Physically Constrained Topology Search To mitigate the combinatorial explosion inherent in reaction network discovery, we formulate

Read the complete text