Skip to content

About

macOS app for running Boltz-2 biomolecular interaction predictions

Topics

Resources

Stars

5 stars

Watchers

0 watching

Forks

Repository files navigation

Multiplexer

Multiplexer

Open-source macOS desktop app for Boltz-2 structure and affinity predictions. Paste SMILES, import a CSV, and get predicted 3D structures back – all on your desktop.


Download for macOS


What it does

Multiplexer makes it easy to run Boltz-2 predictions while working with local files. Paste a target sequence, drop in a list of compounds, and review predicted structures and binding metrics in a single interface. It connects to the Boltz Lab cloud API, which has a generous free tier - you just need to sign up for an API key.

All data stays on your machine. Multiplexer stores everything locally and only sends sequences and SMILES to the Boltz Lab API for prediction.

Getting started

  1. Install - Open the .dmg and drag Multiplexer to Applications.
  2. API key - Go to lab.boltz.bio, create a free account, and generate an API key. Paste it into Settings inside the app.
  3. New campaign - Click "New Campaign", name it, and paste your target sequence.
  4. Add compounds - Click "New Run". Paste SMILES (one per line, or name,SMILES), or import a CSV file.
  5. Submit - Hit Submit. Compounds are sent to Boltz Lab in parallel with automatic rate-limit handling and retries.
  6. Review - Click any completed compound to see the predicted 3D structure, confidence metrics, and 2D structure.
  7. Export - Click "Export CSV" to download all results.

Features

Input

  • Paste SMILES or name,SMILES pairs
  • CSV/TSV import with automatic column detection
  • Real-time SMILES validation via RDKit

Screening

  • Batch concurrent submission (5 at a time)
  • Automatic rate-limit handling and retries for large batches
  • Configurable Boltz-2 model parameters (recycling steps, diffusion samples, step scale)
  • Desktop notifications on run completion

Review

  • Interactive 3D structure viewer (Mol*) with confidence coloring
  • Six metrics per compound: binding confidence, optimization score, structure confidence, pLDDT, ipTM, pTM
  • Sortable results table
  • CSV export

App

  • Uses your own Boltz Lab API key
  • All data stays on your machine
  • Light and dark mode

Tech stack

Electron, React, TypeScript, Tailwind CSS, tRPC, Mol*, RDKit WASM, Zustand

License

MIT

About

macOS app for running Boltz-2 biomolecular interaction predictions

Topics

Resources

Stars

5 stars

Watchers

0 watching

Forks

Releases

Contributors

Languages