Evaluation of optimisation and Bayesian inference methods for reaction rates in atmospheric chemical mechanisms
Authors: Valery Ashu, Wenqing Peng, Zhi-Song Liu, Heikki Haario, Andreas Rupp, Taiwo Ashu, Petri Clusius, Lukas Pichelstorfer, +2 more
Organizations: Department of Computational Engineering, LUT University, 15210, Lahti, Finland · Atmospheric Modelling Center-Lahti, 15140, 15140, Lahti, Finland · Institute for Atmospheric and Earth System Research / Physics, Faculty of Science, University of Helsinki, 00014, Helsinki, Finland · Saarland University, Campus E1.1, 66123, Saarbrücken, Germany · pi-numerics, Neumarkt am W., 5202, Austria
Constraining reaction rate coefficients is a central challenge in the development of explicit atmospheric chemical mechanisms, particularly for autoxidation systems where many reaction pathways are only indirectly observed through high-resolution mass spectrometry. In this study, we evaluate rate-coefficient optimisation methods for a toy-case autoxidation mechanism using synthetic data with known ground truth. Two complementary approaches are compared: ODE-constrained neural-network optimisation, which provides efficient point estimates of uncertain rate coefficients, and the Markov Chain Monte Carlo (MCMC) approach, which samples the posterior distribution of rate coefficients and quantifies parameter uncertainty. The methods are tested using direct concentration observations and mass-spectral observations under different noise levels. For unperturbed and low-noise synthetic observations, both methods converged towards the known rate coefficients, with the neural-network optimiser providing faster point estimates. Under high-noise conditions (with the signal-to-noise ratio approximately S / N = 1), however, MCMC was substantially more robust in recovering the rate coefficients. The posterior analysis shows that mass-spectral aggregation broadens credible intervals even at low noise, and that high-noise mass spectra can leave many individual reaction rates weakly identifiable. Posterior predictive validation nevertheless shows how broad parameter uncertainty constrained by MCMC remains consistent with accurate reproduction of the observable mass spectrum. These results demonstrate that point-estimation and Bayesian sampling methods provide complementary information: neural-network optimisation is effective for informative data, whereas MCMC is essential for diagnosing uncertainty, non-uniqueness, and identifiability in noisy or aggregated inverse problems.
Figures & tables
Measured data
Baseline
noise-free case
low-noise case
high-noise case
MCMC
NN
MCMC
NN
MCMC
NN
Concentration
0.0928
0.0120
0.000147
0.0146
0.0114
0.0768
0.884
Mass spectrum
0.0928
0.0780
0.0402
0.0930
0.0746
0.2921
1.306
Table 1: Overall RMSRE comparison of MCMC and Neural Network across noise levels. The MCMC values are computed from posterior median rate estimates. The baseline RMSRE denotes the value between the initial and true coefficient estimates.
Figure 1: Log relative error of estimated rate coefficients
Figure 2: Optimisation convergence. Panels 2(a) and 2(c) show the trajectory-fitting objective, and panels 2(b) and 2(d) the corresponding error of the estimated rate coefficients. Solid lines use a single noisy realisation; dashed lines use stochastic optimisation over n=4 noisy samples.
Baseline
Low-noise
High-noise
n=1
n=4
n=1
n=4
Conc.
0.0928
0.0114
0.00367
0.884
N/A
Mass
0.0928
0.0746
0.0478
1.306
0.289
Table 2: RMSRE of estimated rate coefficients with additional noise samples.
Figure 3: Comparison of optimisation against ODE trajectory and finite difference derivative, using 0% noise concentration observations. t is the number of time steps sampled from the trajectories.
Case
95% CI width
Median range
Coverage
Concentration
0.015
0.937–1.069
45/50
Mass spectra, low noise
0.047
0.703–1.246
47/50
Mass spectra, high noise
0.613
0.294–2.472
42/50
Table 3: Posterior constraints from the three DRAM-MCMC cases after discarding the first 20 % of each chain as burn-in. The CI width is reported as the median parameter-wise 95 % credible-interval width in log10 multiplier space. The posterior median range is reported in rate-multiplier space. Coverage gives the number of inferred parameters whose 95 % credible interval contains the true multiplier value of one.
Figure 4: Sorted widths of the parameter-wise 95 % credible intervals for the three MCMC cases. Direct concentration observations yield the narrowest posterior intervals. Mass-spectral aggregation broadens the posterior under low-noise conditions, while the high-noise mass-spectral case produces substantially weaker constraints for many parameters.
Figure 5: Posterior medians and 95 % credible intervals for all 50 rate multipliers. The dashed line denotes the true multiplier value of one. The concentration and low-noise mass-spectral cases constrain most parameters tightly around the true value, whereas the high-noise mass-spectral case produces broad intervals and stronger parameter-to-parameter variation in posterior uncertainty.
Figure 6: Representative marginal posterior distributions for selected rate multipliers in log10 space. The vertical dashed line marks the true value. The examples illustrate the transition from sharply localized posteriors in the concentration and low-noise mass-spectral cases to broader and more weakly identified posteriors in the high-noise mass-spectral case.
Figure 7: Posterior predictive check for the high-noise mass-spectral case. The black crosses show the noisy observation, the blue open circles show the clean synthetic spectrum, and the brown markers show the posterior mean prediction from 200 posterior samples. Error bars denote the 90 % posterior predictive interval for each nominal mass channel.
Simulation-based Bayesian inference for Chemical Reaction Networks (CRNs) must simultaneously solve two computationally demanding problems: discriminating between overlapping reaction pathways and estimating reaction rate constants. Classical approaches, such as Sequential Monte Carlo ABC, rely on handcrafted summary statistics and scale poorly across large parameter spaces. To overcome this, we introduce a neural inference architecture. Our pipeline utilizes shared 1D-CNNs and a permutation-invariant Set Attention mechanism to learn low-dimensional summary statistics from variable-length, multi-experiment time-series data. These statistics simultaneously condition a model selection network to compute posterior model probabilities and model-specific conditional Invertible Neural Networks (cINNs) that estimate parameter posterior distributions. We demonstrate the practical utility of our approach on a real protodeboronation example, illustrating how it handles structural ambiguity and extracts kinetic parameters from competing catalyst deactivation pathways.
J. Arturo Esquivel, Jorge A. Campos-Gonzalez-Angulo, Sergio Pablo-García +2
Department of Statistical Sciences, University of Toronto, Toronto, Ontario, Canada · Schools of Mathematics & Physics and Chemistry & Molecular Biosciences, The University of Queensland, St Lucia, Queensland, Australia · Australian Research Council Centre of Excellence in Quantum Biotechnology, St Lucia, Queensland, Australia +9
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 H2 + Br2 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% 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.
Runzhe Liu, Zihao Wang, Wenbo Yang +1
State Key Laboratory of Fine Chemicals, Frontier Science Center for Smart Materials, Dalian University of Technology, Dalian, 116024, China · Dalian Key Laboratory of Intelligent Chemistry, CR Belt and Road Joint Laboratory on Intelligent Chemistry and Advanced Materials of Liaoning Province, School of Chemistry, Dalian University of Technology, Dalian, 116024, China
Chemical reaction-condition optimization -- choosing the catalyst, ligand, solvent, reagent, temperature, time, and atmosphere that jointly maximize yield and stereoselectivity -- is a central, judgement-laden subtask of organic methodology research that large language models are increasingly expected to support. Yet existing chemistry benchmarks evaluate reaction-class labelling, retrosynthesis, or SMILES manipulation, and do not ask models to read a real condition-screening table and pick the best set. We introduce RxnOptBench, a benchmark whose every option and precedent is a real wet-lab entry mined from the optimization tables of organic-methodology papers published in 2025, graded by a continuous relative score derived from a declared headline utility that combines reported yield with enantiomeric excess (ee), diastereomeric ratio (dr), and regioisomeric ratio (rr), and equipped with a paired precedents-vs-no-precedents design that isolates in-context use of literature evidence from parametric memorization. Across nine frontier LLMs and three Chemistry LLMs, even the best models leave substantial headroom: chemistry-specialized models fall to the random-baseline floor on multi-axis selection, while open-weight models have closed most of the gap to proprietary frontier models. We release the final human-reviewed benchmark test set and evaluation code.
Lingli Ge, Yubin Wang, Junyuan Gao +9
Shanghai Jiao Tong University, Shanghai, China · Shanghai Innovation Institute, Shanghai, China · Shanghai Artificial Intelligence Laboratory, Shanghai, China