PharmPapp MCP

MCP tools for cyclic peptide computational analysis - Permeability prediction and molecular descriptor calculation

Table of Contents

• Overview
• Installation
• Local Usage (Scripts)
• MCP Server Installation
• Using with Claude Code
• Using with Gemini CLI
• Available Tools
• Examples
• Troubleshooting

Overview

This MCP (Model Context Protocol) server provides computational tools for cyclic peptide analysis based on the PharmPapp framework. It offers both synchronous and asynchronous APIs for molecular property calculation, permeability prediction, and batch analysis workflows commonly used in drug discovery pipelines.

Features

• Molecular descriptor calculation for cyclic peptides using RDKit
• Permeability prediction using 5 PharmPapp models (RRCK-C, PAMPA-C, Caco2-L/C/A)
• Batch analysis with multi-model comparison and consensus predictions
• Job management for long-running computations with real-time monitoring
• Seamless LLM integration through standardized MCP protocol

Directory Structure

./
├── README.md               # This file
├── env/                    # Conda environment
├── src/
│   └── server.py           # MCP server with 13 tools
├── scripts/
│   ├── calculate_descriptors.py  # Molecular descriptor calculation
│   ├── predict_permeability.py   # Permeability prediction
│   ├── batch_analysis.py         # Multi-model comparison
│   └── lib/                      # Shared utilities (18 functions)
├── examples/
│   └── data/               # Demo data
│       ├── demo_cyclic_peptides.smi     # Sample cyclic peptide SMILES
│       ├── example for Caco2_A.csv      # Model training data (10 molecules)
│       ├── example for Caco2_C.csv      # Model training data (10 molecules)
│       ├── example for Caco2_L.csv      # Model training data (10 molecules)
│       ├── example for PAMPA-C.csv      # Model training data (10 molecules)
│       └── example for RRCK-C.csv       # Model training data (10 molecules)
├── configs/                # Configuration files
│   ├── calculate_descriptors_config.json
│   ├── predict_permeability_config.json
│   ├── batch_analysis_config.json
│   └── default_config.json
└── repo/                   # Original PharmPapp repository

---

Installation

Quick Setup

Run the automated setup script:

./quick_setup.sh

This will create the environment and install all dependencies automatically.

Manual Setup (Advanced)

For manual installation or customization, follow these steps.

Prerequisites

• Conda or Mamba (mamba recommended for faster installation)
• Python 3.10+
• RDKit (installed automatically via conda-forge)

Create Environment

Please follow the procedure from reports/step3_environment.md. The recommended workflow is shown below:

# Navigate to the MCP directory
cd /home/xux/Desktop/CycPepMCP/CycPepMCP/tool-mcps/pharmpapp_mcp

# Check for package manager preference
if command -v mamba &> /dev/null; then
    PKG_MGR="mamba"
else
    PKG_MGR="conda"
fi
echo "Using package manager: $PKG_MGR"

# Create conda environment (use mamba if available)
mamba create -p ./env python=3.10 -y
# or: conda create -p ./env python=3.10 -y

# Activate environment
mamba activate ./env
# or: conda activate ./env

# Install scientific computing dependencies
mamba run -p ./env pip install pandas numpy scikit-learn matplotlib seaborn tqdm loguru click

# Install RDKit from conda-forge (essential for molecular work)
mamba install -p ./env -c conda-forge rdkit -y

# Install MCP dependencies
mamba run -p ./env pip install --force-reinstall --no-cache-dir fastmcp

---

Local Usage (Scripts)

You can use the scripts directly without MCP for local processing.

Available Scripts

Script	Description	Example
scripts/calculate_descriptors.py	Calculate molecular descriptors from SMILES/SDF	See below
scripts/predict_permeability.py	Predict permeability using PharmPapp models	See below
scripts/batch_analysis.py	Batch analysis and model comparison	See below

Script Examples

Calculate Molecular Descriptors

# Activate environment
mamba activate ./env

# Run with demo data
python scripts/calculate_descriptors.py --demo

# Process custom SMILES file
python scripts/calculate_descriptors.py \
  --input molecules.smi \
  --output descriptors.csv

# Include additional descriptor types
python scripts/calculate_descriptors.py \
  --input examples/data/demo_cyclic_peptides.smi \
  --output full_descriptors.csv \
  --config configs/calculate_descriptors_config.json

Parameters:

• --input, -i: Input SMILES (.smi) or SDF (.sdf) file (required)
• --output, -o: Output CSV file path (default: auto-generated)
• --config: JSON configuration file (default: basic descriptors)
• --demo: Use built-in demo data (3 cyclic peptides)

Predict Permeability

python scripts/predict_permeability.py \
  --input descriptors.csv \
  --model caco2_c \
  --output predictions.csv

# Available models: rrck_c, pampa_c, caco2_l, caco2_c, caco2_a
python scripts/predict_permeability.py \
  --input examples/data/"example for Caco2_C.csv" \
  --model caco2_c \
  --demo

Batch Analysis

python scripts/batch_analysis.py \
  --input descriptors.csv \
  --output batch_results/ \
  --models caco2_c rrck_c pampa_c

---

MCP Server Installation

Option 1: Using fastmcp (Recommended)

# Install MCP server for Claude Code
mamba run -p ./env fastmcp install src/server.py --name cycpep-tools

Option 2: Manual Installation for Claude Code

# Add MCP server to Claude Code
claude mcp add cycpep-tools -- $(pwd)/env/bin/python $(pwd)/src/server.py

# Verify installation
claude mcp list

Option 3: Configure in settings.json

Add to ~/.claude/settings.json:

{
  "mcpServers": {
    "cycpep-tools": {
      "command": "/home/xux/Desktop/CycPepMCP/CycPepMCP/tool-mcps/pharmpapp_mcp/env/bin/python",
      "args": ["/home/xux/Desktop/CycPepMCP/CycPepMCP/tool-mcps/pharmpapp_mcp/src/server.py"]
    }
  }
}

---

Using with Claude Code

After installing the MCP server, you can use it directly in Claude Code.

Quick Start

# Start Claude Code
claude

Example Prompts

Tool Discovery

What cyclic peptide tools are available from cycpep-tools?

Property Calculation (Fast)

Calculate molecular descriptors for examples/data/demo_cyclic_peptides.smi and save to descriptors.csv

Permeability Prediction

Predict permeability using the caco2_c model for @examples/data/"example for Caco2_C.csv"

Structure Validation

Validate the input file @examples/data/demo_cyclic_peptides.smi and tell me what's in it

Batch Processing (Submit API)

Submit a batch analysis job for @descriptors.csv using models caco2_c and rrck_c, save results to batch_results/

Job Management

Check the status of job abc12345
Get the logs for job abc12345
List all my jobs

Using @ References

In Claude Code, use @ to reference files and directories:

Reference	Description
@examples/data/demo_cyclic_peptides.smi	Reference demo SMILES file
@configs/calculate_descriptors_config.json	Reference config file
@results/	Reference output directory

---

Using with Gemini CLI

Configuration

Add to ~/.gemini/settings.json:

{
  "mcpServers": {
    "cycpep-tools": {
      "command": "/home/xux/Desktop/CycPepMCP/CycPepMCP/tool-mcps/pharmpapp_mcp/env/bin/python",
      "args": ["/home/xux/Desktop/CycPepMCP/CycPepMCP/tool-mcps/pharmpapp_mcp/src/server.py"]
    }
  }
}

Example Prompts

# Start Gemini CLI
gemini

# Example prompts (same as Claude Code)
> What cyclic peptide tools are available?
> Calculate descriptors for the demo cyclic peptides
> Submit a batch analysis for multiple permeability models

---

Available Tools

Quick Operations (Sync API)

These tools return results immediately (< 10 minutes):

Tool	Description	Parameters
calculate_cyclic_peptide_descriptors	Calculate molecular descriptors	input_file, output_file, include_3d, include_fingerprints
predict_cyclic_peptide_permeability	Predict permeability with single model	input_file, model, output_file, demo
get_available_models	List available permeability models	None
validate_input_file	Validate SMILES/SDF file	input_file
get_example_data	List available example data	None

Long-Running Tasks (Submit API)

These tools return a job_id for tracking (> 10 minutes):

Tool	Description	Parameters
submit_batch_analysis	Multi-model comparison analysis	input_file, output_dir, models, create_plots
submit_large_descriptor_calculation	Large-scale descriptor calculation	input_file, output_file, include_3d, include_fingerprints

Job Management Tools

Tool	Description
get_job_status	Check job progress and runtime
get_job_result	Get results when completed
get_job_log	View execution logs (tail 50)
cancel_job	Cancel running job
list_jobs	List all jobs with status filter
cleanup_old_jobs	Remove old completed jobs

---

Examples

Example 1: Quick Descriptor Calculation

Goal: Calculate basic molecular properties for cyclic peptides

Using Script:

python scripts/calculate_descriptors.py \
  --input examples/data/demo_cyclic_peptides.smi \
  --output demo_descriptors.csv

Using MCP (in Claude Code):

Calculate molecular descriptors for @examples/data/demo_cyclic_peptides.smi and save to demo_descriptors.csv

Expected Output:

• CSV file with 144 molecular descriptors per molecule
• Descriptors include: MolWt, LogP, TPSA, NumHBD, NumHBA, etc.
• Processing time: ~3 seconds for 3 molecules

Example 2: Permeability Prediction

Goal: Predict cyclic peptide permeability using PharmPapp models

Using Script:

python scripts/predict_permeability.py \
  --input "examples/data/example for Caco2_C.csv" \
  --model caco2_c \
  --output caco2_predictions.csv

Using MCP (in Claude Code):

Predict permeability using caco2_c model for the example Caco2_C data and save results

Expected Output:

• Predicted permeability values with confidence scores
• Model performance metrics (R², RMSE)
• Processing time: 1-3 seconds for 10 molecules

Example 3: Multi-Model Comparison

Goal: Compare predictions across different permeability models

Using MCP (in Claude Code):

Submit a batch analysis for @examples/data/"example for Caco2_C.csv" using all available models. Save results to batch_comparison/

Then check the job status and show me the results when complete.

Workflow:

1. Job submitted with unique job_id
2. Monitor progress with get_job_status
3. Retrieve results with get_job_result
4. View detailed logs with get_job_log

Example 4: End-to-End Pipeline

Goal: Complete workflow from SMILES to multi-model predictions

Using MCP (in Claude Code):

1. First, calculate descriptors for @examples/data/demo_cyclic_peptides.smi
2. Then submit a batch analysis using the calculated descriptors with models caco2_c and rrck_c
3. Monitor the job and show me a summary when complete

---

Demo Data

The examples/data/ directory contains sample data for testing:

File	Description	Molecules	Use With
demo_cyclic_peptides.smi	Sample cyclic peptide SMILES	3	descriptor calculation
example for Caco2_A.csv	Caco2-A training data	10	permeability prediction
example for Caco2_C.csv	Caco2-C training data	10	permeability prediction
example for Caco2_L.csv	Caco2-L training data	10	permeability prediction
example for PAMPA-C.csv	PAMPA-C training data	10	permeability prediction
example for RRCK-C.csv	RRCK-C training data	10	permeability prediction

Sample Cyclic Peptide SMILES

The demo file contains 3 representative cyclic peptides:

C1C[C@H](NC(=O)[C@H](NC(=O)[C@H](NC1=O)CC2=CC=CC=C2)CC3=CC=C(C=C3)O)CC(C)C
C[C@H]1NC(=O)[C@H](NC(=O)[C@H](NC(=O)[C@H](NC(=O)[C@H](NC1=O)CC2=CC=CC=C2)CC(C)C)CC3=CC=C(C=C3)O)CC4=CNC5=CC=CC=C54
CC[C@H](C)[C@H]1NC(=O)[C@H](NC(=O)[C@H](NC(=O)[C@H](NC(=O)[C@H](NC1=O)C)CC2=CC=CC=C2)CC(C)C)CC3=CC=C(C=C3)O

---

Configuration Files

The configs/ directory contains configuration templates:

Config	Description	Key Parameters
calculate_descriptors_config.json	Descriptor calculation settings	include_3d, include_fingerprints, descriptor_types
predict_permeability_config.json	Model prediction settings	test_size, cross_validation, scale_features
batch_analysis_config.json	Batch analysis settings	models_to_test, create_plots
default_config.json	General settings	random_state, n_jobs

Configuration Example

{
  "processing": {
    "include_3d": false,
    "include_fingerprints": false,
    "optimize_geometry": false
  },
  "descriptor_types": ["basic", "topological", "constitutional", "physicochemical"],
  "quality_control": {
    "validate_cyclic_peptide": true,
    "calculate_lipinski_violations": true
  }
}

---

Troubleshooting

Environment Issues

Problem: Environment not found

# Recreate environment
mamba create -p ./env python=3.10 -y
mamba activate ./env
mamba run -p ./env pip install pandas numpy scikit-learn matplotlib seaborn tqdm loguru click
mamba install -p ./env -c conda-forge rdkit -y
mamba run -p ./env pip install --force-reinstall --no-cache-dir fastmcp

Problem: RDKit import errors

# Install RDKit from conda-forge (essential)
mamba install -p ./env -c conda-forge rdkit -y

# Test RDKit installation
mamba run -p ./env python -c "from rdkit import Chem; print('RDKit working!')"

Problem: Import errors

# Verify all dependencies
mamba run -p ./env python -c "
from src.server import mcp
from rdkit import Chem
import pandas
import fastmcp
print('All dependencies working!')
"

MCP Issues

Problem: Server not found in Claude Code

# Check MCP registration
claude mcp list

# Re-add if needed
claude mcp remove cycpep-tools
claude mcp add cycpep-tools -- $(pwd)/env/bin/python $(pwd)/src/server.py

Problem: Invalid SMILES error

Ensure your SMILES string is valid for cyclic peptides. Use the cyclo() notation
or ensure ring closure is properly specified with numbers (e.g., C1...C1).

Problem: Tools not working

# Test server directly
mamba run -p ./env python -c "
from src.server import mcp
print('Available tools:', list(mcp.list_tools().keys()))
print('Total tools:', len(mcp.list_tools()))
"

Job Issues

Problem: Job stuck in pending

# Check job directory and logs
ls -la jobs/

In Claude Code:

Get the log for job <job_id> with 100 lines to see what happened

Problem: Descriptor compatibility error**

Error: Too many missing descriptors. Available: 63/94

Solution: This occurs when trying to predict permeability with descriptors that don't match the model requirements. Use the example CSV files for compatible data, or enhance descriptor calculation to include MOE-style descriptors.

Problem: Out of memory errors

Use submit_large_descriptor_calculation for datasets >1000 molecules
Process in smaller batches for very large datasets

Common Model Issues

Problem: "PAMPA model requires 77 descriptors, found 0"

This is a known limitation. The descriptor calculation produces RDKit descriptors while the models expect MOE2D descriptors. Solutions:

1. Use the provided example CSV files which have compatible descriptors
2. Wait for enhanced descriptor mapping in future versions

Debugging

In Claude Code:

# Check job status
Check the status of job <job_id>

# View recent logs
Get the log for job <job_id> with 50 lines

# List all jobs
List all my jobs and their statuses

# Validate input data
Validate the input file @examples/data/demo_cyclic_peptides.smi

---

Development

Running Tests

# Activate environment
mamba activate ./env

# Test basic functionality
python scripts/calculate_descriptors.py --demo

Starting Dev Server

# Run MCP server in development mode
mamba run -p ./env fastmcp dev src/server.py

# Test server connectivity
mamba run -p ./env python -c "from src.server import mcp; print('Server ready')"

---

Performance Guidelines

When to Use Sync vs Submit API

Use Sync API when:

• Dataset < 1000 molecules
• Only basic 2D descriptors needed
• Interactive analysis requiring immediate feedback
• Quick validation and exploration

Use Submit API when:

• Dataset > 1000 molecules
• Including 3D descriptors or fingerprints (much slower)
• Batch processing multiple models
• Long-running analyses that may take >10 minutes

Resource Usage

Operation	Memory	CPU	Time (1000 molecules)
Basic descriptors	~100MB	Low	~2 minutes
3D descriptors	~500MB	High	~15 minutes
Fingerprints	~200MB	Medium	~5 minutes
Model prediction	~50MB	Low	~1 minute
Batch analysis	~300MB	Medium	~10 minutes

---

License

Based on PharmPapp - Fraunhofer ITMP implementation

Credits

This MCP server implements Python equivalents of the KNIME workflows from the original PharmPapp repository, providing programmatic access to cyclic peptide permeability prediction models through modern LLM interfaces.

---

API Reference Summary

Job Management (6 tools)

• get_job_status - Check job progress and runtime
• get_job_result - Get completed job results
• get_job_log - View execution logs
• cancel_job - Cancel running job
• list_jobs - List jobs with status filter
• cleanup_old_jobs - Remove old completed jobs

Synchronous Analysis (2 tools)

• calculate_cyclic_peptide_descriptors - Fast descriptor calculation
• predict_cyclic_peptide_permeability - Single model permeability prediction

Asynchronous Processing (2 tools)

• submit_batch_analysis - Multi-model comparison
• submit_large_descriptor_calculation - Large-scale descriptor calculation

Utilities (3 tools)

• get_available_models - Model information and requirements
• validate_input_file - File format validation
• get_example_data - Available demo data information

Total: 13 MCP tools covering the complete cyclic peptide computational analysis workflow.