Related Concept Videos
You might also read
Related Articles
Articles linked to this work by shared authors, journal, and citation graph.
Activation mechanism of the full-length histidine kinase LvrB from pathogenic Leptospira.
Alpha-synuclein interacts with regulators of ATP homeostasis in mitochondria.
Related Experiment Video
Updated: Sep 8, 2025

Author Spotlight: Enhancing PSC-to-Functional Cell Differentiation Using ML Models Based on Live-Cell Bright-Field Imaging
Published on: October 4, 2024
Predictions of Steady-State Photo-CIDNP Enhancement by Machine Learning.
Marta Stefańska1, Thomas Müntener1, Sebastian Hiller1
1Biozentrum, University of Basel, Spitalstrasse 41, Basel 4056, Switzerland.
Machine learning accurately predicts photochemically induced dynamic nuclear polarization (photo-CIDNP) signal enhancement. This approach enables virtual prescreening of molecules, reducing the need for time-consuming experimental screening in NMR spectroscopy.
More Related Videos
11:49A Continuous-flow Photocatalytic Reactor for the Precisely Controlled Deposition of Metallic Nanoparticles
Published on: April 10, 2019
11:26Integrating a Triplet-triplet Annihilation Up-conversion System to Enhance Dye-sensitized Solar Cell Response to Sub-bandgap Light
Published on: September 12, 2014
Area of Science:
- Nuclear Magnetic Resonance (NMR) spectroscopy
- Hyperpolarization techniques
- Computational chemistry
Background:
- Photochemically induced dynamic nuclear polarization (photo-CIDNP) enhances NMR signal sensitivity.
- Current limitations exist in predicting steady-state photo-CIDNP enhancement, necessitating extensive experimental screening.
- Developing predictive models is crucial for efficient identification of suitable target molecules.
Purpose of the Study:
- To explore the application of machine learning for predicting steady-state photo-CIDNP signal-to-noise enhancement (SNE).
- To identify key molecular features that correlate with photo-CIDNP enhancement.
- To enable virtual prescreening of compound libraries for photo-CIDNP applications.
Main Methods:
- Measurement of steady-state photo-CIDNP SNE for 40 derivatives (indole, amino acid, phenol).
- Correlation of experimental SNE with eight molecular features.
- Development and evaluation of machine learning models (Logistic Regression, CatBoost Regressor, K-Nearest Neighbors).
Main Results:
- The nucleophilic Fukui index was identified as a key qualitative predictor of maximal SNE.
- A Logistic Regression model achieved 100% accuracy in identifying sites with high enhancement (SNE > 90).
- CatBoost Regressor and K-Nearest Neighbors demonstrated superior performance in quantitative SNE prediction.
Conclusions:
- Machine learning models show significant potential for predicting photo-CIDNP SNE.
- This approach can streamline the identification of molecules for hyperpolarization experiments.
- Virtual prescreening using ML can accelerate research in NMR spectroscopy.