autodE is a Python module initially designed for the automated calculation of reaction profiles from SMILES strings of reactant(s) and product(s). Current features include: transition state location, conformer searching, atom mapping, Python wrappers for a range of electronic structure theory codes, SMILES parsing, association complex generation, and reaction profile generation.
Starting with autodE v1.4.5 (gomesgroup fork), ORCA 6.x is the recommended high-level QM backend for automated reaction discovery. Here's why:
| Feature | Benefit for autodE | Alternative |
|---|---|---|
| GOAT (Global Optimizer with Adaptive Thermostat) | Automated conformer search with metadynamics - finds global minima without manual intervention | CREST (requires separate setup) |
| SOLVATOR | Automatic explicit solvation shell generation - critical for accurate solution-phase barriers | Manual placement |
| DOCKER | Built-in molecular docking for bimolecular reactions - automatic association complex generation | External docking tools |
| NEB-TS | Improved nudged elastic band for complex reaction paths | Standard NEB |
| IRC | Enhanced intrinsic reaction coordinate with bidirectional tracing | Manual IRC setup |
| ExtOpt | External optimizer interface - enables MLIP-accelerated geometry optimization | Pure DFT (slower) |
| Free & Open | No license fees, unlimited parallel jobs | Commercial alternatives |
GOAT (Metadynamics-Based Conformer Search)
import autode as ade
from autode.wrappers.keywords import GOATKeywords
# Automatic global minimum search
mol = ade.Molecule(smiles='CC(C)CC(C)C') # 2,4-dimethylpentane
mol.optimise(method=ade.methods.ORCA(), keywords=GOATKeywords())
# GOAT automatically explores conformational spaceSOLVATOR (Explicit Solvation)
from autode.wrappers.keywords import SolvatorKeywords
# Add explicit water molecules around solute
keywords = SolvatorKeywords(solvent='water', n_solvent=50, shell_radius=5.0)
solvated_mol = mol.optimise(method=ade.methods.ORCA(), keywords=keywords)MLIP-Accelerated Optimization
from autode.wrappers import MLIPConfig
# Use machine learning potentials for fast pre-optimization
config = MLIPConfig(model='aimnet2', server_url='http://localhost:5003')
mol.optimise(method=ade.methods.ORCA(), mlip_config=config)
# 10-100x faster initial optimization before DFT refinementDOCKER (Bimolecular Complex Generation)
from autode.wrappers import dock_molecules
# Automatic docking for reaction complex setup
reactant1 = ade.Molecule(smiles='CCBr')
reactant2 = ade.Molecule(smiles='[OH-]')
complex = dock_molecules(reactant1, reactant2, method=ade.methods.ORCA())
# Finds optimal pre-reactive geometry for SN2 reactionThis fork fixes a critical bug where ORCA 6.x was incorrectly detected as ORCA 4.x:
# Before (broken): version[0] == "5" returns False for "6.1.1"
# After (fixed): int(version.split('.')[0]) >= 5 returns True
# New version properties available:
orca = ade.methods.ORCA()
print(orca.is_v6) # True for ORCA 6.x
print(orca.is_v5_or_later) # True for ORCA 5.x and 6.x
print(orca.major_version) # 6- Python >= 3.9
- Recommended High-Level Method:
- ORCA >= 6.0 (Recommended - free, full feature support)
- Alternative High-Level Methods:
- ORCA 4.x-5.x (limited features)
- Gaussian09
- Gaussian16
- NWChem >= 6.5
- QChem >= 5.4
- GPU4PySCF
- Recommended Low-Level Method:
- XTB >= 6.5 (Recommended)
- Alternative Low-Level Methods:
Optional Dependencies:
- orca-pi - ORCA Python Interface for enhanced parsing
- MLIP API Server - For MLIP-accelerated optimization
To install autodE with conda:
conda install autode -c conda-forge
To install the gomesgroup fork with ORCA 6.x features:
pip install git+https://github.com/gomesgroup/autode.git
See the installation guide for installing from source.
Reaction profiles in autodE are generated by initialising Reactant and Product objects, generating a Reaction from those and invoking calculate_reaction_profile(). For example, to calculate the profile for a 1,2 hydrogen shift in a propyl radical:
import autode as ade
ade.Config.n_cores = 8
ade.Config.hcode = "orca" # Recommended: ORCA 6.x
ade.Config.lcode = "xtb"
r = ade.Reactant(name='reactant', smiles='CC[C]([H])[H]')
p = ade.Product(name='product', smiles='C[C]([H])C')
reaction = ade.Reaction(r, p, name='1-2_shift')
reaction.calculate_reaction_profile() # creates 1-2_shift/ and saves profileModern DFT with D4 Dispersion:
import autode as ade
from autode.wrappers.keywords.orca6 import orca6_keywords, r2scan3c_keywords
ade.Config.hcode = "orca"
# Use r2SCAN-3c (fast, accurate composite method)
mol = ade.Molecule(smiles='c1ccccc1')
mol.optimise(method=ade.methods.ORCA(), keywords=r2scan3c_keywords)IRC Validation of Transition States:
from autode.wrappers import run_irc, validate_ts_with_irc
# Find TS
ts = reaction.find_ts()
# Validate with IRC (traces path both directions)
irc_result = validate_ts_with_irc(ts, method=ade.methods.ORCA())
if irc_result.connects_to_reactants and irc_result.connects_to_products:
print("TS validated!")Hybrid MLIP-DFT NEB:
from autode.wrappers.neb import NEBMLIPHybrid
# Fast NEB with MLIP, refined with DFT at key points
neb = NEBMLIPHybrid(
reactant=r, product=p,
mlip_model='aimnet2',
dft_method=ade.methods.ORCA()
)
ts_guess = neb.get_ts_guess()See examples/ for more examples and duartegroup.github.io/autodE/ for additional documentation.
| Feature | Module | Description |
|---|---|---|
| GOAT | autode.wrappers.keywords.goat |
Metadynamics conformer search |
| SOLVATOR | autode.wrappers.keywords.solvator |
Explicit solvation shell generation |
| DOCKER | autode.wrappers.docker |
Molecular docking for complexes |
| IRC | autode.wrappers.irc |
Intrinsic reaction coordinate |
| NEB-TS | autode.wrappers.keywords.neb |
Enhanced nudged elastic band |
| MLIP | autode.wrappers.mlip_external |
Machine learning potential integration |
| OPI | autode.wrappers.opi_wrapper |
ORCA Python Interface |
| D4 | autode.wrappers.keywords.dispersion |
D4 dispersion correction |
| r2SCAN-3c | autode.wrappers.keywords.orca6 |
Composite DFT method |
| Relativistic | autode.wrappers.keywords.relativistic |
ZORA, X2C, DKH2 |
For detailed documentation on ORCA 6.x features, see docs/orca6_features.md.
The gomesgroup fork supports multiple architectures for MLIP integration:
| Architecture | Platform | MLIP Server |
|---|---|---|
| x86_64 | Linux (Intel/AMD) | localhost:5003 |
| aarch64 | Linux (ARM64/Grace Hopper) | localhost:5003 |
| arm64 + Darwin | macOS (Apple Silicon) | localhost:5003 |
There is a slack workspace for development and discussion - please email to be added. Pull requests are very welcome but must pass all the unit tests prior to being merged. Please write code and tests! See the todo list for features on the horizon. Bugs and feature requests should be raised on the issue page.
NOTE: We'd love more contributors to this project!
# Run all tests
pytest tests/
# Run ORCA 6.x specific tests
pytest tests/test_orca_v6*.py -v
# Run with coverage
pytest tests/ --cov=autode --cov-report=htmlIf autodE is used in a publication please consider citing the paper:
@article{autodE,
doi = {10.1002/anie.202011941},
url = {https://doi.org/10.1002/anie.202011941},
year = {2021},
publisher = {Wiley},
volume = {60},
number = {8},
pages = {4266--4274},
author = {Tom A. Young and Joseph J. Silcock and Alistair J. Sterling and Fernanda Duarte},
title = {{autodE}: Automated Calculation of Reaction Energy Profiles -- Application to Organic and Organometallic Reactions},
journal = {Angewandte Chemie International Edition}
}
- Tom Young (@t-young31)
- Joseph Silcock (@josephsilcock)
- Kjell Jorner (@kjelljorner)
- Thibault Lestang (@tlestang)
- Domen Pregeljc (@dpregeljc)
- Jonathon Vandezande (@jevandezande)
- Shoubhik Maiti (@shoubhikraj)
- Daniel Hollas (@danielhollas)
- Nils Heunemann (@nilsheunemann)
- Sijie Fu (@sijiefu)
- Javier Alfonso (@javialra97)
- Robert MacKnight (@rmacknight99)
- Jose Regio (@jregio)
- Gabe Gomes (@gabegomes) - gomesgroup fork maintainer