Deeptech · AI-driven Bioinformatics & Drug Discovery Bengaluru, India
info@silicoscientia.com LinkedIn
Request a demo
Partners · Collaboration

Boltz models, inside SilicoXplore.

Through our collaboration with Boltz, SilicoXplore integrates Boltz models next to its thirty one modules. Structure and affinity prediction, protein design and small-molecule hit discovery are reached from the same guided forms as docking, dynamics and free energy, and their results carry straight into each other.

See SilicoXplore Visit Boltz
About Boltz

A family of open models for biomolecular design.

Boltz is a family of biomolecular AI models. The first releases, Boltz-1 and Boltz-2, grew out of research at the MIT Jameel Clinic, and Boltz PBC now continues the work, describing its aim as “advancing the open frontier of biomolecular AI”.

Boltz-2 was the first model of its kind to predict a complex’s structure and the binding affinity of its ligand together, and BoltzGen extended the family to designing protein binders. Both are open source. BoltzMol-1, BoltzProt-1 and the Boltz API followed, for small-molecule hit discovery, protein design and programmatic access.

Developer
Boltz PBC, with roots in research at the MIT Jameel Clinic
Models
Boltz-2, BoltzGen, BoltzMol-1, BoltzProt-1 and the Boltz API
Open source
Boltz-2 and BoltzGen are released as open source
Our work
SilicoXplore integrates Boltz models through our collaboration
Website
boltz.com
What is in SilicoXplore

The Boltz packages,
in one workspace.

Boltz’s five packages, used alongside the thirty one SilicoXplore modules, from the same guided forms and without any programming.

Structure and affinity

Boltz-2

Predicts the three dimensional structure of a complex together with the binding affinity of its ligand.

Open source

Protein design

BoltzGen

Generates new protein binders for a target.

Open source

Small molecules

BoltzMol-1

Boltz’s pipeline for small-molecule screening and hit discovery.

Protein design

BoltzProt-1

Boltz’s protein design pipeline, a step on from BoltzGen.

Access

Boltz API

Boltz’s interface for running its models inside pipelines, which SilicoXplore makes available alongside its own modules.

Boltz-2 and BoltzGen are released as open source. BoltzMol-1, BoltzProt-1 and the Boltz API are used under Boltz’s own terms.

How the collaboration works

From a Boltz result to a ranked candidate.

Because Boltz outputs enter the same chemical representation as every other SilicoXplore module, a predicted structure or a generated binder passes straight into the next step.

  1. 1One place to start

    Boltz models are reached from the same guided forms as the other thirty one modules.

  2. 2Predict or design

    Predict a complex and its affinity, design a protein binder, or search for small-molecule hits.

  3. 3Dock and simulate

    Carry the result into docking and pose evaluation, then into dynamics and free energy.

  4. 4Decide with evidence

    Compare the AI prediction with physics-based estimates before anything goes to the bench.

Boltz is a separate organisation. Its models are developed by Boltz and used under their own terms. Computational results remain hypotheses that require experimental validation. For the models themselves, see boltz.com.

See it on your target

Run Boltz next to docking, dynamics and free energy.

Tell us the target and the question, and we will show how the Boltz models and the SilicoXplore modules answer it together.

Request a demo Explore the modules