Skip to main content
OSTI.GOVU.S. Department of Energy Office of Scientific and Technical Information
U.S. Department of Energy
Office of Scientific and Technical Information
 

Uncertainty quantification for molecular property predictions with graph neural architecture search

Journal Article · · Digital Discovery
DOI:https://doi.org/10.1039/d4dd00088a· OSTI ID:2438089
 [1];  [1];  [1];  [2];  [3]
  1. University of Wisconsin, Madison, WI (United States)
  2. Oak Ridge National Laboratory (ORNL), Oak Ridge, TN (United States)
  3. University of Wisconsin, Madison, WI (United States); Argonne National Laboratory (ANL), Argonne, IL (United States)
+ Show Author Affiliations
Graph Neural Networks (GNNs) have emerged as a prominent class of data-driven methods for molecular property prediction. However, a key limitation of typical GNN models is their inability to quantify uncertainties in the predictions. This capability is crucial for ensuring the trustworthy use and deployment of models in downstream tasks. To that end, we introduce AutoGNNUQ, an automated uncertainty quantification (UQ) approach for molecular property prediction. AutoGNNUQ leverages architecture search to generate an ensemble of high-performing GNNs, enabling the estimation of predictive uncertainties. Our approach employs variance decomposition to separate data (aleatoric) and model (epistemic) uncertainties, providing valuable insights for reducing them. In our computational experiments, we demonstrate that AutoGNNUQ outperforms existing UQ methods in terms of both prediction accuracy and UQ performance on multiple benchmark datasets, and generalizes well to out-of-distribution datasets. Additionally, we utilize t-SNE visualization to explore correlations between molecular features and uncertainty, offering insight for dataset improvement. AutoGNNUQ has broad applicability in domains such as drug discovery and materials science, where accurate uncertainty quantification is crucial for decision-making.
Research Organization:
Argonne National Laboratory (ANL), Argonne, IL (United States)
Sponsoring Organization:
National Science Foundation (NSF); USDOE; USDOE Office of Science (SC), Advanced Scientific Computing Research (ASCR)
Grant/Contract Number:
AC02-06CH11357
OSTI ID:
2438089
Alternate ID(s):
OSTI ID: 2377227
OSTI ID: 2477241
Journal Information:
Digital Discovery, Journal Name: Digital Discovery Journal Issue: 8 Vol. 3; ISSN 2635-098X
Publisher:
Royal Society of ChemistryCopyright Statement
Country of Publication:
United States
Language:
English

References (42)

Ensembles of Networks Produced from Neural Architecture Search book January 2020
Machine Learning book January 2021
FreeSolv: a database of experimental and calculated hydration free energies, with input files journal June 2014
On ensemble techniques of weight-constrained neural networks journal January 2020
Uncertainty quantification in scientific machine learning: Methods, metrics, and comparisons journal March 2023
Applications of Deep Learning in Molecule Generation and Molecular Property Prediction journal December 2020
Uncertainty Quantification Using Neural Networks for Molecular Property Prediction journal July 2020
Evaluating Scalable Uncertainty Estimation Methods for Deep Learning-Based Molecular Property Prediction journal April 2020
Evidential Deep Learning for Guided Molecular Property Prediction and Discovery journal July 2021
PotentialNet for Molecular Property Prediction journal November 2018
Reoptimization of MDL Keys for Use in Drug Discovery journal November 2002
ESOL:  Estimating Aqueous Solubility Directly from Molecular Structure journal May 2004
Enumeration of 166 Billion Organic Small Molecules in the Chemical Universe Database GDB-17 journal November 2012
970 Million Druglike Small Molecules for Virtual Screening in the Chemical Universe Database GDB-13 journal July 2009
QSAR Modeling: Where Have You Been? Where Are You Going To? journal January 2014
Quantum chemistry structures and properties of 134 kilo molecules journal August 2014
MoleculeNet: a benchmark for molecular machine learning journal January 2018
A quantitative uncertainty metric controls error in neural network-driven chemical discovery journal January 2019
Capturing molecular interactions in graph neural networks: a case study in multi-component phase equilibrium journal January 2023
Prediction uncertainty validation for computational chemists journal October 2022
Machine learning of molecular electronic properties in chemical compound space journal September 2013
ChEMBL: towards direct deposition of bioassay data journal November 2018
Neural network ensembles journal January 1990
Graph Neural Network Architecture Search for Molecular Property Prediction conference December 2020
DeepHyper: Asynchronous Hyperparameter Search for Deep Neural Networks conference December 2018
Estimating the mean and variance of the target probability distribution conference January 1994
Graph Neural Networks in TensorFlow and Keras with Spektral [Application Notes] journal February 2021
ASGN: An Active Semi-supervised Graph Neural Network for Molecular Property Prediction conference August 2020
Monte Carlo Dropout Based BatchEnsemble For Improving Uncertainty Estimation
  • Jain, Shubham; P. K., Srijith
  • Proceedings of the 6th Joint International Conference on Data Science & Management of Data (10th ACM IKDD CODS and 28th COMAD) https://doi.org/10.1145/3570991.3571038
conference January 2023
Dataset’s chemical diversity limits the generalizability of machine learning predictions journal November 2019
Explainable uncertainty quantifications for deep learning-based molecular property prediction journal February 2023
Extended study on atomic featurization in graph neural networks for molecular property prediction journal September 2023
Evaluating uncertainty-based active learning for accelerating the generalization of molecular property prediction journal November 2023
An integrative machine learning approach for prediction of toxicity-related drug safety journal November 2018
Adam: A Method for Stochastic Optimization preprint January 2014
Bayesian Convolutional Neural Networks with Bernoulli Approximate Variational Inference preprint January 2015
Graph Attention Networks preprint January 2017
How Powerful are Graph Neural Networks? preprint January 2018
Recalibration of Aleatoric and Epistemic Regression Uncertainty in Medical Imaging preprint January 2021
A Survey of Uncertainty in Deep Neural Networks preprint January 2021
Uncertainty Toolbox: an Open-Source Library for Assessing, Visualizing, and Improving Uncertainty Quantification preprint January 2021
AutoDEUQ: Automated Deep Ensemble with Uncertainty Quantification preprint January 2021

Similar Records

Clarifying trust of materials property predictions using neural networks with distribution-specific uncertainty quantification
Journal Article · Tue May 02 20:00:00 EDT 2023 · Machine Learning: Science and Technology · OSTI ID:2422892

Related Subjects

37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CHEMISTRY
97 MATHEMATICS AND COMPUTING