Primer on Model Context Protocol for Biotech Researchers

Published on May 19, 2025 • 7 min read

Imagine asking an AI assistant to analyze the latest protein structure data, cross-reference it with clinical trial results, and then pull up relevant antibody reagents - all in a single conversation AND with the confidence that its response will be based on real data pulled in from specified databases. It's what the Model Context Protocol (MCP) makes possible today. For biotech researchers and biologists, MCP represents a fundamental shift in how we interact with scientific data and AI tools.

The Challenge: Data Silos in Biotech Research

As biotech researchers, we work with an ever-expanding universe of data sources: protein databases like UniProt, genomic repositories like NCBI, structural data from PDB, clinical trial registries, literature databases, and proprietary lab systems. Each has its own interface, API, and access method. When you need to answer a complex research question, you often find yourself:

  • Switching between dozens of different websites and databases
  • Computational scientists learning multiple API specifications and authentication methods
  • Writing custom scripts to integrate data from different sources
  • Manually copying and pasting data between tools
  • Struggling to keep AI assistants connected to real-time, authoritative data while wondering if the model hallucinated and just making up responses.

This fragmentation slows down research and creates barriers to discovery. MCP can make all this a little easier.

What is Model Context Protocol?

Model Context Protocol (MCP) is an open standard developed by Anthropic that enables AI applications to securely connect to data sources and tools through a unified interface. Think of it as a universal adapter that allows AI assistants like Claude to seamlessly access and interact with external systems.

Instead of building separate integrations for every database and tool, MCP provides a standardized way for:

  • Data Access: AI systems to query databases, read files, and fetch information from APIs
  • Tool Usage: AI assistants to execute functions, run analyses, and perform computations
  • Context Sharing: Applications to provide relevant information to AI models in a structured way

Simple Analogy

If AI models are like brilliant research assistants, MCP is the key that unlocks the laboratory doors, the library, and the data archives - giving your AI assistant secure access to all the knowledge and tools it needs to help you with your research.

How Does MCP Work?

MCP operates through a client-server architecture with three key components:

MCP Hosts (AI Applications)

These are the AI applications you interact with, like UbiChat, Claude Desktop, ChatGPT, etc. They initiate connections to access data and tools.

MCP Servers (Data & Tool Providers)

These are lightweight programs that expose specific data sources or tools to AI applications. For example, an MCP server might provide access to UniProt, PubMed, or your internal LIMS system.

MCP Protocol (The Standard)

A standardized communication protocol that defines how hosts and servers talk to each other, ensuring compatibility and security across different implementations.

What MCP Servers Can Provide

MCP servers can expose three types of capabilities to AI applications:

  • Resources - Data that AI can read, like database records, files, or API responses (e.g., protein sequences from UniProt, publications from PubMed)
  • Prompts - Pre-built templates for common tasks (e.g., "analyze this protein sequence for functional domains" or "summarize these clinical trial results")
  • Tools - Functions AI can execute (e.g., BLAST searches, molecular docking simulations, statistical analyses)

The Universal Translator for AI

With powerful tools like Boltz2 and intelligent LLMs, how do we make them talk to each other? This diagram explains the role of an MCP Server, the crucial "universal adapter" that enables seamless communication and makes complex science accessible.

🤖

LLM Assistant

e.g., UbiChat, chatGPT, Claude

↓
🔌

MCP Server

The "Universal Adapter"

↓
🔬

Scientific Tool

e.g., Immunebuilder, Boltz2

Why MCP Matters for Biotech Research

MCP addresses critical challenges that biotech researchers face daily when working with AI and data:

Truth-Grounded AI

AI assistants can query authoritative databases directly instead of relying on potentially outdated training data. This reduces hallucinations and ensures responses are backed by current, real data.

Unified Access

Connect to multiple biomedical databases through a single, standardized interface. No more juggling different APIs, authentication methods, or data formats.

Reproducibility

MCP servers document exactly which data sources and versions were used, making research more reproducible and transparent.

Privacy & Security

Keep sensitive data local. MCP servers run on your infrastructure, so proprietary data never leaves your environment unless you explicitly allow it.

Real-World Examples in Biotech

Here's how MCP is already transforming biotech research workflows:

Target Identification & Validation

An MCP server connected to Open Targets enables you to ask: "What genetic evidence supports BRAF as a therapeutic target for melanoma?"

The AI can instantly query Open Targets' comprehensive database, retrieve GWAS data, somatic mutations, clinical evidence, and provide a synthesized validation report—all without you leaving your AI chat interface.

Literature Review & Meta-Analysis

With MCP servers for PubMed and bioRxiv, you can request: "Find all studies from the last 5 years on CRISPR-based treatments for sickle cell disease and summarize the clinical outcomes."

The AI retrieves current publications, extracts relevant data, and provides a comprehensive summary with proper citations—saving hours of manual literature review.

Protein Structure Analysis

An MCP server for AlphaFold or Boltz-2 allows queries like: "Predict the structure of this novel fusion protein and identify potential binding pockets."

The AI can execute structure prediction, analyze the results, and suggest druggable sites—accelerating the early stages of drug design.

Experimental Planning

With an Antibody Registry MCP server: "Find validated antibodies for Western blot detection of phosphorylated STAT3."

The AI retrieves antibodies with verified applications, citations, and vendor information—ensuring reproducible experimental design.

Getting Started with MCP

The MCP ecosystem is rapidly growing. Here's how you can start using it in your research:

1. Choose an MCP-Compatible AI Platform

  • Claude Desktop: Native MCP support from Anthropic
  • ChatGPT: Integration capabilities for custom GPTs
  • UbiChat: Our biotech-focused platform with pre-configured MCP servers for drug discovery

2. Build and Install Custom Servers (Optional)

We provide MCP servers for many commonly used databases. For proprietary data or internal tools, you can develop custom MCP servers. The protocol is open-source and well-documented, with SDKs available in Python, TypeScript, and other languages.

UniBio Intelligence MCP Servers

We've developed production-ready MCP servers specifically for biotech research, including Open Targets, Antibody Registry, and Boltz-2 for structure prediction. These servers are optimized for drug discovery workflows and come with comprehensive documentation.

Interested in using our MCP servers? Reach out at contact@unibiointelligence.com to learn more.

Want to Learn More?

Explore our other blog posts to see MCP in action:

← Back to Blog