PhAMP-KG: Agentic RAG & Knowledge Graph for AMP Discovery

🚧 Status: Active Development / Under Construction 🚧

Overview

PhAMP-KG (Antimicrobial Peptide Knowledge Graph) is a production-grade pipeline designed to mine PubMed literature for multidrug-resistant (MDR) pathogens and antimicrobial peptides (AMPs). The system leverages ESM-C to encode peptide sequences and constructs a robust Neo4j knowledge graph. By employing a LangChain ReAct agent, PhAMP-KG facilitates hypothesis generation and orchestrates discovery downstream of experimentally validated leads.

Architecture

%% Placeholder for Architecture Diagram
graph TD
    A[PubMed Literature] --> B[fetch_literature.py]
    B --> C[extract_entities.py]
    C --> D[Neo4j Ingestion]
    E[AMP Sequences] --> F[embed_sequences.py]
    F --> D
    D --> G[Graph Representation Learning]
    H[LangChain ReAct Agent] --> D
    H --> I[Hypothesis Generation]

Loading

Clarivate JD Mapping

Industry Requirement	PhAMP-KG Implementation
NLP pipelines	SciSpaCy + UMLS EntityLinker
Knowledge Graph Construction	Neo4j + Cypher
Graph Representation Learning	Node2Vec + ESM-C hybrid vector
Reproducible Workflows	Nextflow DSL2 + Docker Compose
Agentic RAG	LangChain ReAct Agent

Quickstart

Docker Compose

docker-compose up -d

Nextflow Pipeline

nextflow run main.nf