Intelligene Search | AI-Powered Genomics Search Engine

Best Data Science Project Award - Agentic search engine for genomics research

Intelligene Search - “Perplexity for Genomics”

  • Best Data Science Project Award at Bilkent University (out of 44 teams)
  • Built an agentic search engine for genomics that combines Graph RAG, Semantic RAG, and Deep Graph Infomax
  • Processes 300+ million nodes and 1+ billion edges from GenomicKB
  • Three intelligent search modes powered by LLMs and custom query generation

Why We Built This

Biology research is drowning in data. Researchers spend countless hours trying to find connections in genomic information, and existing tools just aren’t cutting it. We built Intelligene to be different - an AI knowledge engine that actually understands genomics and gives clear answers to complex questions.

What Makes It Special

Unlike traditional search tools, Intelligene works like having a genomics expert at your fingertips. It combines three powerful AI approaches:

  • Graph RAG: Uses our custom Cypher query engine to extract relationships from the knowledge graph
  • Semantic RAG: Finds semantically related information using vector embeddings
  • Deep Graph Infomax (DGI): Discovers hidden connections that other methods miss

Three Modes for Different Needs

Reasoning Mode

Perfect for quick queries. Processes your question through all three AI paths simultaneously and combines the results for comprehensive answers.

Agentic Mode

For complex questions, this mode uses a ReAct agent that dynamically chooses which tools to use, iteratively refining results. Includes 70+ integrations with tools like Slack and Notion.

Deep Research Mode

When you need serious analysis. Conducts 20+ minute deep dives across the entire knowledge graph and textual data for the most demanding research questions.

The Technical Side

The entire system is deployed live on our servers, handling:

  • GenomicKB dataset with 300+ million nodes
  • Real-time query processing across 1+ billion edges
  • Custom Cypher query generation
  • Multi-modal AI reasoning

The Team

This was a team effort with Yusuf Kesmen, Mehmet Bahattin Dulkadiroğlu, Alpsencer Özdemir, and Yiğit Küçük, with support from Furkan Özden (Oxford), under the guidance of Prof. A. Ercument Cicek at Bilkent.

Recognition

We’re grateful for the support from:

  • Lidya Genomics and Ciceklab for collaboration
  • Kuika Software and Türenç Engin for the award
  • Mert Bicakci for semester-long support
  • Bilkent University and all our professors

Making genomic research faster, smarter, and more accessible - one query at a time.