Dimitri StudioStructure prediction, local

RNA · protein · DNA on Apple Silicon

Fold proteins and RNA on your Mac.

Pick a model, paste a sequence, watch it fold on the Apple GPU, then inspect, compare and export the structure. Your sequences and results never leave the Mac.

Download for macOS ▶ Watch the launch film

Apple Silicon (M1 or later) · macOS 14 or later · version 0.2.0

The launch film

Sixty-two seconds on what changed: AlphaFold-class prediction that runs on the laptop in front of you.

2.4 s

Warm folds

RhoFold+ keeps the model loaded between jobs. The 52-nt 6TB7 RNA folds in a median 2.4 s, 2.4× faster than a cold process.

100%

On the Mac

Single-sequence mode: no MSA search, no templates, nothing uploaded. The viewer, models and results all run offline.

3

Models

RhoFold+ for RNA, AlphaFold 3 through mac-af3, and OpenFold3-MLX. Install each from the Library with one click.

4

Exports

mmCIF, PDB, JSON and a PNG of the view. Drop an mmCIF or PDB back in to open it as a result.

Inside the app

A 3D cartoon coloured by pLDDT, a per-residue confidence plot, a PAE heatmap and the models compared on the same sequence.

Results screen: a RhoFold+ prediction coloured by pLDDT
Results — pLDDT, PAE and RNA secondary structure for one prediction.
Predict screen with a sequence and model picker
Predict in light mode — paste a sequence or drop a FASTA; protein, RNA, DNA or ligand is detected.
Compare screen with superposed structures and RMSD
Compare — superposition, RMSD and per-residue deviation across models.
Library screen listing installed models
Library — install models, see what is loaded and what is on disk.

Models

Each model runs in its own worker process. The first launch sets up the app's Python environment by itself, with uv bundled inside the app.

ModelRuntimeDeviceInputsConfidence
RhoFold+rhofold-mac (PyTorch)Apple GPU (MPS), CPU fallbackRNApLDDT, secondary structure
AlphaFold 3mac-af3 on alphafold3-mac (MLX)Apple GPU (MLX)protein, RNA, DNA, ligandpLDDT, pTM, ipTM, PAE
OpenFold3-MLXopenfold-3-mlxMPSprotein, RNA, DNApLDDT, pTM, ipTM

Swipe the table sideways to see every column.

Download Dimitri Studio

Version 0.2.0, build 33, arm64. 28 MB disk image.

Download the disk image SHA-256

Signed, not yet notarized. The app carries a Developer ID signature and the hardened runtime, but Apple has not notarized it yet. On first open macOS may say it cannot check the app for malicious software. Control-click the app, choose Open, then Open again.

SHA-256 58e94b5ef6946c830776d2015ff5f80c1fb36d41de29344f6c8906ab04e75397