openbio

What This Skill Does

The OpenBio skill provides an integrated gateway to global biological databases and advanced computational biology tools. Designed for AI agents, it bridges the gap between raw data retrieval and complex analysis by enabling programmatic access to PDB, UniProt, ChEMBL, PubMed, and high-performance structure prediction pipelines like Boltz and ProteinMPNN. It functions as a laboratory-in-the-loop, allowing for precise molecular biology experiment design, sequence analysis, and clinical variant interpretation without manual database browsing.

Installation

To integrate OpenBio into your environment, use the OpenClaw CLI:

clawhub install openclaw/skills/skills/ravishar313/openbio

Ensure you have configured your authentication by setting the OPENBIO_API_KEY environment variable in your system path or .env file to authorize requests to the https://openbio-api.fly.dev/ base URL.

Use Cases

• Structural Biology: Predict protein folds or analyze existing PDB entries for binding site viability.
• Drug Discovery: Query ChEMBL and PubChem to evaluate small molecule inhibitors or calculate Tanimoto similarity thresholds.
• Scientific Literature: Automate the discovery of relevant research papers by filtering PubMed and arXiv based on current project hypotheses.
• Genomics: Perform variant annotation and assess statistical significance using GWAS data thresholds.
• Molecular Design: Automatically generate primers or simulate restriction enzyme cloning strategies.

Example Prompts

1. "OpenBio, find all papers published in the last month on PubMed regarding CRISPR-Cas9 genome editing in human T-cells and summarize the key findings."
2. "Use the Boltz tool to predict the tertiary structure of the provided protein sequence and report the pLDDT confidence scores for the active site."
3. "Analyze this variant list against the ClinVar database and filter for pathogenic mutations with a p-value less than 5e-8."

Tips & Limitations

• Schema First: Always invoke the tool schema lookup before calling a function, as parameter naming conventions (e.g., pdb_ids) are strict and subject to update.
• Job Management: Long-running computational tasks like structure prediction return a job_id. Monitor status using the /jobs/{job_id}/status endpoint before requesting final output.
• Confidence Matters: When reviewing prediction data, apply the skill's quality thresholds: treat pLDDT scores < 50 as disordered and prioritize structural resolution < 2.5 Å for molecular docking simulations.