Skip to main content
Esenbil Information Technologies
Esenbil Information Technologies

AgnostiMap: Proteome-Wide Interaction Discovery
Eliminate off-target blindspots by mapping where your molecule sequesters in the human proteome

About the Project

AgnostiMap is a target-agnostic drug–proteome mapping platform designed to reveal how small molecules interact with the full human proteome. Using AlphaFold 3 (v6) derived structures, the system screens candidate ligands against more than 20,000 proteins without prior target assumptions.


AgnostiMap operates through a rigorous, multi-stage pipeline:

  • High-Throughput Screening: Systematic docking estimations across the entire proteome to identify potential binding sites.

  • MD Validation:Molecular dynamics simulations (10 ns+) to test binding persistence and cavity stability under realistic motion.

  • AI-Driven Fusion: Fusion of structural and energetic signals into interpretable maps of sequestration risk, off-target exposure, and intracellular sink formation.


By replacing narrow safety panels with a global interaction map, AgnostiMap exposes hidden binding reservoirs that drive hepatic accumulation and unpredictable clinical toxicity ← the true blind spots of modern drug discovery.


Platform Specifications

  • Proteome Coverage: 20,000+ human proteins (AlphaFold v3–derived structures)

  • Ligand Scope: < 800 Da (small-molecule optimized)

  • Dynamic Validation: Molecular dynamics simulations using OpenMM


  • End-to-End Docking Workflow

    Pipeline Overview

    Stage 1

  • Initial docking affinity estimations using known protein structures.

  • Integrates protein data from the AlphaFold v3 Human Proteome.

  • Physical and chemical sanity check
  • AI based false positive elimination
  • Stage 2

  • In-depth molecular dynamics simulations including:

    1. Protein/Ligand Preparation

    2. System Assembly

    3. Solvation & Ionization

    4. Energy Minimization

    5. Equilibration

    6. Full Simulation

  • Configured via AI Assistant to tailor simulation parameters.

  • Stage 3

  • Outputs from dynamic analysis are integrated using a Mixture of Experts ensemble model
  • Stage 4

  • Reviewed by a combined AI Assistant + Human Expert team for final binding affinity classification and druggability decision.
  • Samples

    Ligand: Sebetralstat
    P03952
    P07148
    P41247

    Technology Stack

  • 🧠 AI: Google Gemini & OpeanAI ChatGPT

  • 🧪 Molecular Simulation: OpenMM for dynamics

  • 🧬 Structure Data: AlphaFold v3 human proteome, UniProt, STRING, Open Targets
  • ☁️ Infrastructure: Google Cloud Platform and Vertex AI (Grant-supported)

  • Collaboration

    In collaboration with:

    • Yeditepe University Medical School

    • Google Cloud for Research (Cloud Credit Grant)

    Contact

    🔹 Interested in industry collaboration?

    📧 Email us