Official Resources
- Source Repository: https://github.com/patonlab/CASCADE
- Documentation: Included in repository
- License: Open source
Overview
CASCADE (CAlculation of NMR Chemical Shifts using DEep learning) is a tool for predicting NMR chemical shifts using machine learning. It combines DFT-calculated shifts with ML corrections to achieve CCSD(T)-quality predictions at DFT cost, particularly for organic molecules.
Scientific domain: NMR chemical shift prediction, ML-accelerated spectroscopy
Target user community: Researchers predicting NMR chemical shifts for organic molecules with ML-enhanced accuracy
Theoretical Methods
- DFT NMR chemical shift calculation
- Machine learning correction (Δ-ML)
- CCSD(T) quality from DFT+ML
- 1H and 13C chemical shift prediction
- Spin-orbit relativistic corrections (ΔSO)
Capabilities (CRITICAL)
- NMR chemical shift prediction
- ML correction of DFT shifts
- 1H and 13C chemical shifts
- CCSD(T)-quality accuracy at DFT cost
- Spin-orbit relativistic corrections
- Organic molecule support
Sources: GitHub repository, J. Chem. Inf. Model.
Key Strengths
ML-Enhanced Accuracy:
- CCSD(T)-quality predictions
- DFT cost with ML correction
- Systematic improvement
- Trained on high-level data
NMR-Specific:
- 1H and 13C chemical shifts
- Spin-orbit corrections
- Heavy atom effects
- Organic molecule focus
DFT Integration:
- Uses DFT as baseline
- ML correction on top
- Any DFT functional as input
- Flexible framework
Inputs & Outputs
-
Input formats:
- Molecular geometry
- DFT-calculated shifts
- ML model files
-
Output data types:
- Corrected chemical shifts
- ML correction values
- Comparison with DFT
- Prediction accuracy
Interfaces & Ecosystem
- DFT codes: Baseline shift calculation
- PyTorch/scikit-learn: ML backend
- Python: Core language
Performance Characteristics
- Speed: Fast (ML prediction)
- Accuracy: CCSD(T)-quality
- System size: Organic molecules
- Memory: Low
Computational Cost
- ML prediction: Seconds
- DFT baseline: Minutes (separate)
- Typical: Very efficient
Limitations & Known Constraints
- Organic molecules: Limited to organic systems
- 1H and 13C only: No other nuclei
- Training data dependent: Quality limited by training set
- DFT baseline needed: Requires DFT calculation first
Comparison with Other Codes
- vs ShiftML: CASCADE is organic-focused, ShiftML is solid-state NMR
- vs ml4nmr: CASCADE is Δ-ML correction, ml4nmr is direct ML correction
- vs afnmr: CASCADE is ML, afnmr is bio-NMR specific
- Unique strength: ML-corrected NMR chemical shifts to CCSD(T) quality for organic molecules
Application Areas
Organic Chemistry:
- Structure verification
- NMR shift prediction
- Isomer identification
- Reaction monitoring
Drug Discovery:
- Small molecule NMR
- Metabolite identification
- Purity assessment
- Structural elucidation
Method Development:
- ML-NMR benchmarking
- DFT functional comparison
- Chemical shift databases
- Accuracy improvement
Best Practices
DFT Baseline:
- Use consistent DFT functional
- Adequate basis set for NMR
- Include solvent effects
- Validate against experiment
ML Correction:
- Use appropriate ML model
- Check domain of applicability
- Validate on test set
- Compare DFT vs corrected
Community and Support
- Open source on GitHub
- Developed by Paton Lab (CSU)
- Published in J. Chem. Inf. Model.
- Research code
Verification & Sources
Primary sources:
- GitHub: https://github.com/patonlab/CASCADE
Confidence: VERIFIED
Verification status: ✅ VERIFIED
- Source code: ACCESSIBLE (GitHub)
- Documentation: Included in repository
- Published methodology: J. Chem. Inf. Model.
- Specialized strength: ML-corrected NMR chemical shifts to CCSD(T) quality for organic molecules