PocketScout MCP

Overview

PocketScout MCP is a binding-site triage server that supplies tools and workflow prompts to an MCP-compatible AI assistant. Its role is to assemble evidence about a protein target before deeper drug-discovery or computational binder-design work. It connects to UniProt, RCSB PDB, ChEMBL, PubMed, and AlphaFold DB; it is a separate integration, not an upstream database service or a database-endorsed implementation.

Inputs include PDB identifiers, UniProt identifiers, gene names, and selected binding-site residues, depending on the tool. The documented workflow establishes biological context and structural coverage, extracts ligand-contact residues from experimental coordinates, reviews ligand bioactivity history, and checks conservation and known variants. Tools return raw information alongside an interpretation field. The target_briefing prompt supports a first-pass summary, while binding_site_assessment guides a more detailed, ranked assessment with evidence, trade-offs, and design considerations.

The structural analysis uses gemmi to identify contacts around co-crystallized ligands and consolidate pockets across structures by recurrence. Conservation checks cover mouse, rat, and cynomolgus macaque using local-context matching rather than full multiple sequence alignment. These outputs can support prioritization, but they are not experimental confirmation of binding or suitability for design.

The package is marked Alpha in its project metadata. Source code and a hosted MCP endpoint are documented, with local operation requiring Python 3.11 or later. Scope is limited to public data and known binding-site evidence: computational pocket prediction, proprietary data access, patent integration, and ensemble-based allosteric detection are not documented as current capabilities.

Key Features

• Eight MCP tools cover target characterization, related structures, binding sites, ligand history, conservation, literature, known variants, and cross-structure pocket consolidation.
• Coordinate-based analysis uses gemmi to extract residues contacting co-crystallized ligands and classify pockets, including a size-based druggability assessment.
• Cross-structure consolidation combines observed pockets for a target and ranks them by recurrence.
• Binding-residue conservation checks compare human sequences with mouse, rat, and cynomolgus macaque through local-context matching; a separate tool flags documented disease or resistance variants.
• Tools provide raw data plus scientific interpretation, with `target_briefing` and `binding_site_assessment` prompts organizing short or detailed assessments.

Use Cases

• Suggested evaluation: compare known binding regions before selecting a pocket for a computational protein-binder design campaign.
• Suggested evaluation: brief a researcher on an unfamiliar target using structural coverage, ligand history, and relevant literature.
• Suggested evaluation: screen candidate binding residues for cross-species differences and documented variants before planning follow-up experiments.

How to Use

1. Read the official README to choose between hosted access and local operation, and distinguish the short briefing from the detailed assessment workflow.
2. For hosted use, open Claude connectors, name the connector PocketScout, and supply https://pocketscout-mcp.up.railway.app/mcp. This connects an assistant to the server; it does not install the underlying databases.
3. For local use, follow the README's package or source installation instructions. Check the Python requirement in pyproject.toml, then use the documented local client configuration.
4. Start with a target and a PDB identifier. Request target_briefing for reconnaissance or binding_site_assessment for a ranked workup; inspect the returned identifiers and residue selections before further analysis.
5. Evaluate recommendations against the cited structures, bioactivity records, and papers. Treat recurrence, conservation, and variant flags as prioritization evidence, and use separate computational or experimental methods to investigate novel pockets or validate binding.