Home page Documentation Video demos
Design protein backbones, run structure and sequence prediction with an integrated, visual-first pipeline.
Molchanica is a molecule visualization, editing, and computation tool. It combines functionality found in PyMol, Chimera, Coot, VMD, GROMACS, and Avogadro, and operates with the focus on visualizing molecules over time and space. Any molecule you open or build can move using molecular dynamics. It integrates with machine-learning tools for protein engineering, including RFDiffusion3, OpenFold3, Boltz-2, OpenDDE, ProteinMPNN, and Chai-1.
Molchanica includes Computer Aided Drug Design (CADD) tools. For example, molecule screening and generation based on pharmacophores, machine-learning from empirical binding data and more. It estimates ADME and toxicity parameters using a neural network trained on Therapeutic data commons (TDC) data.
Molchanica is intuitive enough that formal documentation isn't required to get started. As a reference, see the documentation.
Download from the releases page.
- Windows: Run the
*_setup.exeinstaller. - Linux: Unzip, and run
setup_molchanica.sh(Withoutsudo). - Mac: Unzip and run
molchanica.
The "nocuda" downloads leave out the cuFFT library, only relevant for the Dynamics MD backend. The normal versions will work even if you don't have CUDA or an Nvidia GPU.
Compile from source by downloading and installing Rust, then running cargo build --release from a CLI in the project directory.
Allows one-click installation of external tools. Click the Tools button in the UI to get started. See the docs page for details, and a full list of tools
- GROMACS
- xTB
- Dynamics (native)
- Backbone design using RFDiffusion3
- Structure prediction using OpenFold3, OpenDDE, Boltz-2, Chai-1, Protenix, and EsmFold2
- Sequence prediction using ProteinMPNN and LigandMPNN
- ORCA
- Psi4
- RDKit
- Gemmi
- View the structure of proteins and small molecules
- Create figures for publication
- Pymol-like selection syntax
- Molecular dynamics, using Amber, OPLS, or GFN-FF force fields, and the OPC water model
- Run GROMACS MD, and view trajectories using the same GUI as our native MD
- Run ORCA and Psi4 quantum chemistry procedures on molecules (if ORCA is installed)
- Run structure prediction with OpenFold3, OpenDDE, Boltz-2, and Chai-1
- Run sequence prediction with ProteinMPNN and LigandMPNN
- Scan every point mutation at every position for stability (ΔΔG), in a single pass, natively — no Python needed at run time
- Assess protein-ligand binding (affinity, pose validation, binding spot identification etc)
- Antibody workflows: CDR selection, paratope contacts, and developability triage
- Read a local PDBbind release: look up a complex's measured binding affinity and open its protein, pocket, and ligand files
- Assess pharmacokinetics, screen small molecules for binding to pockets, pharmacophore features, molecule alignment, and more
- Visualize electron density from crystallography and Cryo-Em data
- Screen molecules based on binding affinity and desired properties
- Pharmacophore based modeling and screening
See Bio Files for details; we use it as a dependency, and keep it in sync.
- Proteins: mmCIF
- Small molecules: SDF, Mol2, Mol, GRO, XYZ, and PDBQT
- Electron density: 2fo-fc mmCIF, Map, and MTZ
- Force field parameters: dat, lib, frcmod, prmtop (Amber), and top (GROMACS)
- MD: GRO, MDP, DCD, TRR, XTC
Most functionality works without internet connectivity. If you have internet, it includes API integrations. For example:
- Load molecules and sequences automatically from PubChem, UniProt, drugbank, and RCSB PDB.
- Improved structure prediction accuracy using MSA servers
- Reactions from Rhea
- OPLS MD parameterization for small molecules
- Buttons and links to open relevant database pages for molecules operating on
- Finds data on the molecule you're building or editing as you work, if it's on PubChem.
Use of GPU vs CPU as the computation device depends on which functionality is used. External tools like GROMACS, xTB, RFdiffusion3, ProteinMPNN, and OpenDDE each have their own requirements regarding which GPU vendors (Nvidia, AMD, Apple) are supported. Some programs like ORCA and Psi4 run primarily on CPU.
If an Nvidia GPU of at least RTX 3 series is available, molecular dynamics via the Dynamics backend, and electron density calculations will be performed using the GPU via CUDA kernels.
For CPU computations, we split execution across CPU cores using thread pools, and use SIMD instructions. It uses all cores available, and either 512-bit, or 256-bit SIMD instructions, depending on CPU capability. External tools written in C++ tend to use an MP library (e.g. OpenMP) to perform this thread pooling.
This application integrates classical molecular dynamics setup and playback into the GUI. For details, see the documentation. It can use its [native backend](Dynamics rust library), GROMACS, or xTB, and can automatically install those backends. It uses AMBER, CHARMM, OPLS, or GFN-FF force fields, with automatic parameterization. The Dynamics readme is a good reference for that backend, and MD in general. Reference the [GROMACS Manual] and xTB documentation for those backends respectively.
It supports running MD directly on the following molecule types:
- Proteins and peptides
- Small organic molecules
- DNA and RNA
- Lipids
- Carbohydrates
It can also perform quantum molecular dynamics using ORCA. This is much slower. See this documentation page for details.
Molchanica can play back molecular dynamics trajectories run from other programs, including GROMACS, OpenMM, and NAMD.
We provide an editor for small organic molecules. This is fully-3D, and has integrated dynamics, which can be run continuously while you edit. It ensures the molecules you create are realistic, and take the appropriate shape. See the documentation for details on how the editor works.
Press the "M" key to move the current molecule, or the "R" key to rotate. (Or use the associated UI buttons). From here, you can drag the molecule around on the plane of the screen, or rotate it. You can also use the syntax "m x 2" to move 2 angstroms on the X axis. You can also click and drag arrows drawn in the molecule to confine movement or rotation to a specific axis, and click plane icons to move it along different planes. Controls here are inspired by both PyMol's and Blender's.
You can measure dihedral angle, 3-atom angle, and distances using the Measurements window. These are labeled,a and you can change their displayed color. These are overlayed on the atoms, and displayed in a list.
The camera is set up with 6 degrees of freedom, using either keyboard + mouse, or mouse only. This makes it easy to get any view of the system you want.
There are two camera control schemes, selectable using buttons in the camera section of the GUI.
The free camera mode is intended to be used with a keyboard and mouse together. They operate on the perspective of the viewer, vice the molecule. You can move and rotate and move the camera in 6 degrees of freedom, allowing you to easily view the molecule from any perspective.
Similar to traditional molecular viewing software. The camera arcs (or orbits) around the molecule, when holding the left mouse button and dragging. Other controls, like scroll wheel and middle mouse, operate similar to the free camera.
If orbit sel is set in the GUI, the orbit center will be the selected atom or residue, vice the molecule center.
Most of this functionality is also available in the GUI, and you can find hotkeys from the GUI's tooltips. We recommend starting with the GUI, and integrating hotkeys into your workflow as you go, using tooltips. See below for a reference.
- F3: Pulls up a search for UI items; an alternative to clicking around the UI, especially for items you are not sure of the location.
- Hold the left mouse button while dragging to rotate the camera in pitch and yaw.
- Hold the middle mouse button while dragging to move the camera left, right, up, and down.
- Scroll to move the camera forward and backwards.
- Scroll while holding left mouse button to roll.
- Right click to select the atom or residue under the cursor. This also selects the molecule to manipulate.
-
Tab: Outside the molecule editor, reset to the Front view of the active molecule.
-
W: Move forward
-
A: Move right
-
A: Move left
-
D: Move back
-
Space: Move up
-
C: Move down
-
Q: Roll counter-clockwise
-
R: Roll clockwise
-
Shift (left): Hold to increase camera movement and rotation speed.
-
Scroll whlie holding left mouse: Roll (Alternative to Q/R)
-
Left arrow: select previous residue
-
Right arrow: select next residue
-
Left arrow: Select previous residue
-
Right arrow: Select next residue
-
Left backet: Previous view mode (sticks, surface mesh etc)
-
Right bracket: Next view mode
-
Semicolon: View/select level (Atoms, bonds, residues)
-
Right bracket: Next view/select level
-
M: Move a molecule with the mouse and scroll wheel
-
R: Rotate a molecule with the mouse and scroll wheel
-
Enter: Move the camera to the selected atom or residue.
-
Esc: Exit move/rotate mode if in one. Otherwise, clear the selection (keeping the molecule active); if nothing is selected, deactivate the active molecule.
-
Ctrl + scroll: Change the fog distance. (hide distant atoms)
-
Delete: In the primary mode, close the active molecule. In the molecule editor, this deletes the selected atom or bond.
-
Enter: Move the camera to the selected atom, bond, or residue.
-
Tab: In the molecule editor, add an atom.
Note: The molecule editor has a different set of hot keys; see its tooltips for details.
Supports volumetric and isosurface views for electron density data, e.g. from Cryo-EM and X-Ray crystallography data.
It can download this data from RCSB PDB, or load files directly. To open 2fo-fc and MTZ files, we use the Gemmi program. Install it from the Tools window, like other external tools.
It can import Map files directly, and save load density to Map format.
Uses neural networks to infer pharmacokinetic properties of arbitrary small molecules, using Therapeutic Data Commons data. See the docs for details
See the docs for details
Molchanica supports a very limited subset of PyMol's CLI interface. Supported commands:
help: Lists commandspwdlscdset seq_view
fetch: Loads a protein from the RCSB PDB. e.g.fetch 1C8Ksave: Save the opened protein or small molecule to disk. e.g.save molecules/1htm.cifload: Load a protein or small molecule from disk. e.g.load ../1htm.cif
show representation, expression: Add a representation layer, e.g.show sticks, ligand. Without an expression, applies toall.show_as: Set the global default view mode, e.g.show_as cartoon. Existing layers remain independent.view: Save and load scenes. e.g.view v1,view v1 store,view v2 recallhide: Limited options available, e.g.resn HOH,hydro,chain,hetatmetc.remove: Limited options available, e.g.resn HOH,hydro,chain,hetatmetc.
select expression: Select atoms across loaded molecules.selealso works as the command name.select name, expression: Save or update a named selection.select %name: Recall a named selection. Bare names also work.delete name: Delete an unused named selection. References from other selections or layers must be removed first.
Supports most selection commands from Pymol, e.g.:
select site, byres (protein within 4 of ligand)
show cartoon, protein
show sticks, site and not hydrogen
show spheres, ligand
show surface, protein
It also supports building selection commands via the GUI.
turnmoveorientreset
Using a dropdown in the UI, you can change selection mode between Atom, (covalent) Bond, and Residue. These are mostly self-explanatory. When selecting a component using right click, or toggling using the UI buttons or arrow keys, this changes what is selected, and the nature of information displayed. For example, in Bond mode, you can see bond length, elements of the atoms connected, and the expected oscillation frequency of the bond. In Atom mode, you can see dihedral angles, the atom name as part of its residue, the atom's force field name, and its partial charge.
You can add DNA, RNA, and lipids in various configurations without loading files; this program can create them procedurally using the GUI. It can create DNA and RNA from a given nucleic acid or amino acid sequence. It can create lipids arranged freely, as membrances, or as lipid nanoparticles (LNPs).
molchanica_prefs.mca is a small binary file containing application state. It's what lets the program remember the last file opened, current view settings etc. It will grow with the number of molecules you've opened, as it stores per-molecule settings. Deleting it is harmless, other than resetting these conveniences.
Molchanica keeps its data in one folder: the one its executable is in.
This application is written in rust; it compiles normally using cargo build --release, which produces a standalone executable.
If you're not compiling on a machine with an Nvidia GPU or without the CUDA toolkit installed, append the --no-default-features to the build command. This will disable native-functionality GPU support on the executable produced. (e.g. MD with the Dynamics backend; does not effect external tools)
The bundled molecule databases are optional. If hmdb.parquet and ChEBI.parquet are present in the project root, they're embedded in the executable.
At any given point, Cargo.toml will likely point to path dependency patch overrides, making it difficult to compile. This is due to me updating underlying libraries, in between their publish cycles. My intent is that you use Molchanica by downloading a binary on the releases page, with the source code as a reference. Or, use the underlying libraries (bio_files, bio_tools, na_seq, bio_apis, dynamics etc) as dependencies in your projects.
If compiling with GPU support, your compiling PC must have the CUDA Toolkit, v13+ installed. You must set the environment var LD_LIBARARY_PATH (Linux) or Path (Windows) to your CUDA bin directory, e.g. C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v13.0\bin. You may also need the build tools containing cl.exe or similar in the path, e.g.: C:\Program Files\Microsoft Visual Studio\2022\Community\VC\Tools\MSVC\14.44.35207\bin\Hostx64\x64 for Windows, or apt install build-essential for linux.











