MolSimplify

Overview

molSimplify is a research toolkit developed by the Kulik Group in MIT’s Department of Chemical Engineering for constructing inorganic molecules and intermolecular complexes. Its workflow role is to generate candidate structures for automated first-principles screening and discovery. It combines structure-building capabilities with bundled predictive models, rather than functioning solely as a molecular-property predictor.

Structure generation can begin with a central atom and coordinating ligands, including monodentate and multidentate ligands. Alternatively, an existing metal–ligand structure, such as a porphyrin, can serve as the starting point for adding or replacing ligands. The resulting coordination complexes provide candidates for computational investigation. The toolkit also builds intermolecular complexes intended for studies of binding interactions and for preparing candidate reactants and intermediates in catalyst mechanism screening. The source excerpts establish these input concepts and outputs, but do not specify complete input schemas or generated file formats.

The bundled neural networks predict metal–ligand bond lengths, spin-splitting energies, frontier orbital energies, spin-state-dependent reaction energies, and simulation outcomes for octahedral transition metal complexes. That stated scope should not be generalized to arbitrary molecular classes; the excerpts supply neither accuracy figures nor detailed applicability criteria. Structure generation is supported through the command line and the Python entry point startgen_pythonic; the GUI is no longer supported. Installation guidance offers several routes, but contains inconsistent Python-version examples and notes architecture-dependent availability of optional packages, making environment selection an important preparation step.

Key Features

• Builds metal coordination complexes using monodentate or multidentate ligands around a central atom.
• Functionalizes existing metal–ligand structures by adding ligands or replacing those already present.
• Generates intermolecular complexes for binding-interaction studies and catalyst reactant or intermediate preparation.
• Includes neural networks for bond lengths, spin-splitting energies, frontier orbital energies, spin-state-dependent reaction energies, and simulation outcomes in octahedral transition metal complexes.
• Supports structure generation through the command line or the Python entry point `startgen_pythonic`.

Use Cases

• Suggested evaluation: generate a small set of coordination complexes with different ligand denticities and inspect their suitability as candidates for first-principles screening.
• Suggested evaluation: add or exchange ligands on an existing porphyrin or other metal–ligand complex to explore structural variants.
• Suggested evaluation: prepare intermolecular arrangements representing binding partners, catalyst reactants, or intermediates for subsequent computational investigation.
• Suggested evaluation: assess the bundled property predictors on representative octahedral transition metal complexes using independently established reference values.

How to Use

1. Read the documentation and select a structure-generation example from the Kulik Group tutorials that matches your coordination or intermolecular-complex task.
2. Choose an installation route from the repository README. It describes PyPI, GitHub, conda, and Docker options. Resolve the mismatch between its Python 3.10 recommendation and older environment examples before following them; check optional dependencies for your architecture.
3. Follow the selected installation instructions and run the documented installation tests. Treat results as a local environment check, noting that the README says some tests may be skipped when optional dependencies are absent.
4. Prepare the central atom and ligands, or an existing complex to modify, using the chosen tutorial’s input conventions. Generate structures through the command line or startgen_pythonic, not the unsupported GUI.
5. As an intended evaluation, inspect generated structures before downstream calculations and compare any model predictions with suitable references. Consult the ML model reference page for the relevant scientific citations.