Uncertainty Calibration in Molecular Machine Learning: Comparing Evidential and Ensemble Approaches

2026/02/04 by Bidhan Chandra Garain, Max Pinheiro, Matheus de Oliveira Bispo +1 · 1 voice
Biochemistry, Genetics and Molecular Biology · Computer Science · Materials Science · #Computational Drug Discovery Methods #Machine Learning in Materials Science #Protein Structure and Dynamics

paper · pdf · doi:10.1002/chem.202503299

openalex publication_date 2026/02/04 · openalex created_date 2026/02/06 · openalex updated_date 2026/07/22

Abstract

Machine learning (ML) models are increasingly used in quantum chemistry, but their reliability hinges on uncertainty quantification (UQ). In this study, we compare two prominent UQ paradigms-deep evidential regression (DER) and deep ensembles-on the QM9 and WS22 datasets, with a specific emphasis on the role of post hoc calibration. Raw uncertainties from both methods were systematically miscalibrated: DER produced uncertainty estimates where data noise and model uncertainty were not cleanly separated, while ensembles produced sharper yet underconfident estimates. Applying calibration techniques such as isotonic regression (ISR), standard scaling, and GP-Normal corrected these deficiencies, aligning predicted variances with observed errors. On QM9, calibration enabled DER to filter high-confidence predictions more effectively than ensembles. On WS22, calibrated ensembles not only improved statistical reliability but also delivered substantial computational savings in active learning, reducing redundant ab initio evaluations by more than 20%. These results demonstrate that post hoc calibration is essential to transform uncertainty estimates from descriptive metrics into actionable signals, ensuring both trustworthy predictions and resource-efficient molecular modeling.

Citations

• A few useful things to know about machine learning
• Los carnavales de La Bañeza: Aspectos antropológicos
• A review of uncertainty quantification in deep learning: Techniques, applications and challenges
• Combining Machine Learning and Computational Chemistry for Predictive Insights Into Chemical Systems
• LFaB: Low fidelity as Bias for Active Learning in the chemical configuration space
• Charting electronic-state manifolds across molecules with multi-state learning and gap-driven dynamics via efficient and robust active learning
• Evidential Deep Learning for Interatomic Potentials
• Outlier-Detection for Reactive Machine Learned Potential Energy Surfaces
• Graph Neural Network Interatomic Potential Ensembles with Calibrated Aleatoric and Epistemic Uncertainty on Energy and Forces
• Uncertainty Driven Active Learning of Coarse Grained Free Energy Models
• Uncertainty quantification for predictions of atomistic neural networks
• Uncertainty-aware molecular dynamics from Bayesian active learning for Phase Transformations and Thermal Transport in SiC
• Uncertainty Toolbox: an Open-Source Library for Assessing, Visualizing, and Improving Uncertainty Quantification
• Calibrated Uncertainty for Molecular Property Prediction using Ensembles of Message Passing Neural Networks
• E(n) Equivariant Graph Neural Networks
• Ab initio machine learning in chemical compound space
• Individual Calibration with Randomized Forecasting
• Thousands of reactants and transition states for competing E2 and\n S_\N2 reactions
• Verified Uncertainty Calibration
• Evaluating and Calibrating Uncertainty Prediction in Regression Tasks
• Distribution Calibration for Regression
• Accurate Uncertainties for Deep Learning Using Calibrated Regression
• Simple and Scalable Predictive Uncertainty Estimation using Deep Ensembles
• Dropout as a Bayesian Approximation: Representing Model Uncertainty in Deep Learning

Discussions

• Uncertainty Calibration in Molecular Machine Learning: Comparing Evidential and Ensemble Approaches - Chem Eur J #CompChem doi.org/10.1002/chem... [bsky, 3 points, 0 comments]

Related

• MUBen: Benchmarking the Uncertainty of Molecular Representation Models
• Calibrated Uncertainty for Molecular Property Prediction using Ensembles of Message Passing Neural Networks
• Uncertainty quantification by direct propagation of shallow ensembles
• Evaluating Scalable Uncertainty Estimation Methods for DNN-Based Molecular Property Prediction
• DeepUQ: Assessing the Aleatoric Uncertainties from two Deep Learning Methods