tamarind

Tamarind Bio — Cloud-Based Molecular Design and Structure Prediction Platform

Skills Overview

Tamarind Bio is a cloud computing biology platform that provides API access to open-source tools for protein structure prediction, molecular design, antibody development, and more. Advanced models such as AlphaFold, Boltz-2, and Chai-1 can run without local GPUs.

Use Cases

Protein Structure Prediction and Research: Predict the three-dimensional structures of monomeric proteins, protein complexes, or protein–ligand complexes, including high-accuracy structural modeling using the latest models such as AlphaFold, Boltz-2, and Chai-1.

Protein and Antibody Molecular Design: Perform de novo protein design, binder design, and sequence optimization for antibodies or nanobodies. Use rational design tools such as RFdiffusion and ProteinMPNN.

Batch Screening and Drug Discovery: Process large numbers of sequences in bulk for structural characterization, virtual screening, binding affinity prediction, or protein–ligand docking. Suitable for drug discovery and optimization workflows.

Core Features

Integration of Multiple Structure Prediction Models: The platform integrates mainstream protein structure prediction tools including AlphaFold, Boltz-2, Chai-1, and ESMFold. It supports structure prediction for monomers, multimers, and protein–small-molecule complexes, with no need for users to install or configure local compute environments.

Protein Design and Optimization Toolchain: Provides protein design tools such as RFdiffusion, ProteinMPNN, and BoltzGen. Supports an end-to-end design workflow from backbone generation and sequence design to inverse folding. It also includes dedicated antibody and nanobody design as well as developability evaluation capabilities.

Molecular Docking and Binding Affinity Prediction: Integrates docking tools such as DiffDock and AutoDock Vina to perform protein–small-molecule docking and binding affinity prediction. It also supports molecular dynamics simulations, providing key computational support for drug discovery.

Frequently Asked Questions

Is Tamarind Bio free?

Tamarind Bio provides each user with 10 free job credits, which can be used to try out the platform’s features. For larger-scale usage, contact the Tamarind Bio team to purchase a subscription.

How do I get a Tamarind API key?

Go to app.tamarind.bio, register an account, and create an API key in Account Settings or API Settings. After obtaining the key, set it as the environment variable TAMARIND_API_KEY, or pass it via the x-api-key request header on each API call.

Which protein structure prediction models does Tamarind support?

The platform supports multiple mainstream protein structure prediction models, including AlphaFold (AF2), Boltz-2, Chai-1, and ESMFold/ESMFold2. These models fit different scenarios: AlphaFold supports monomer and multimer prediction; Boltz-2 and Chai-1 support protein–small-molecule complexes; ESMFold provides fast single-sequence prediction. You can view the complete tool list via the /tools endpoint or MCP getAvailableTools.

Can I submit protein design tasks in batches?

Yes. Tamarind provides the /submit-batch endpoint and the MCP submitBatch tool, allowing you to submit up to 100 jobs of the same type at once. This is useful for batch screening of sequences, high-throughput structural characterization, or parallel design tasks. Batch jobs create a parent job; you need to monitor the batchStatus state to wait for all child jobs to complete and for results to be aggregated.

Does Tamarind require a local GPU?

No. Tamarind is a cloud platform, and all compute tasks run on hosted GPUs. Users only need to submit tasks via the REST API or an MCP server, without configuring local hardware environments. This enables researchers without GPU resources to use state-of-the-art computational biology tools.

How do I call the Tamarind API in Python?

Tamarind has no official Python SDK, so you must use the requests library to call the REST API directly. The basic workflow is: set the x-api-key request header, call /tools to discover available tools, get the parameter schema for the chosen tool, submit tasks via /submit-job or /submit-batch, poll the /jobs endpoint to monitor status, and finally download results via /result. For a detailed example, refer to the platform’s OpenAPI specification (app.tamarind.bio/openapi.yaml).

Does Tamarind support antibody and nanobody design?

Yes. The platform provides dedicated antibody and nanobody design tools, including sequence generation, humanization, developability assessment, immunogenicity prediction, and more. Users can use these tools to optimize antibody engineering candidates and evaluate the physicochemical and biological properties of potential drug molecules.

What tools are used for protein–ligand docking?

Tamarind integrates molecular docking tools such as DiffDock and AutoDock Vina. DiffDock is deep-learning-based and is suitable for fast virtual screening. AutoDock Vina is a classic docking tool, suitable for more detailed docking studies. In addition, structure prediction models such as Boltz-2 and Chai-1 also support co-folding prediction for protein–small-molecule complexes, allowing simultaneous prediction of ligand binding sites and protein conformational changes.