Research
The Science Behind Inferrence
Neuro-symbolic conformational exploration reveals drug targets that static structure prediction cannot see.
Beyond Static Structures
AlphaFold revolutionized structural biology by predicting a single, static protein structure with remarkable accuracy. But proteins are not static. They breathe, flex, and transition between conformational states on timescales from picoseconds to seconds.
Cryptic binding sites are pockets that only appear in these transient conformational states. They are invisible in crystal structures and absent from single-structure predictions. Yet they represent some of the most promising drug targets in modern pharmacology.
The Problem with One Structure
- Static predictions miss allosteric pockets entirely
- Crystal structures capture the lowest-energy state, not the druggable state
- MD simulations are computationally prohibitive for screening campaigns
- Inferrence generates hundreds of diverse conformations in minutes, not days
Our Pipeline
From amino acid sequence to drug candidate discovery in minutes, powered by proprietary AI models.
AA Sequence
Submit your amino acid sequence or upload a FASTA file
AI Models
Proprietary deep learning models generate hundreds of diverse 3D conformations from sequence alone
Pocket Analysis
Automated detection and cross-state comparison reveals cryptic and conserved binding sites
Discover Ligands
AI-powered screening ranks candidate compounds for each discovered pocket
What Makes a Site Cryptic?
Invisible in Crystals
Cryptic sites do not exist in the dominant conformational state captured by X-ray crystallography. The pocket is literally absent in the solved structure.
Conformationally Gated
These sites only appear when the protein undergoes specific conformational transitions: loop movements, helix rotations, or domain rearrangements that transiently expose a druggable cavity.
Therapeutically Valuable
Because cryptic sites are allosteric and structurally distinct from the active site, drugs binding here avoid resistance mutations and can modulate function through novel mechanisms.
Ligand Discovery
Finding a cryptic pocket is only half the battle. Inferrence takes every discovered binding site and screens it against compound libraries using AI-powered molecular docking.
For each pocket, our ligand discovery pipeline ranks candidate molecules by predicted binding affinity, drug-likeness, and selectivity — giving you actionable hit compounds, not just structural data.
From target to lead in a single platform. No need to export PDB files and run separate docking campaigns.
What You Get
- Ranked ligand candidates per pocket with binding scores
- Radar charts showing pocket druggability, hydrophobicity, and depth
- ADMET property predictions for top candidates
- Export-ready reports for medicinal chemistry teams
Validated Results
100% success rate on the CryptoSite benchmark. Every known cryptic binding site was recovered from sequence alone.
p38 MAP kinase — DFG-out allosteric site
Key cancer and inflammation target. The DFG-out pocket is invisible in the active conformation and only appears during kinase conformational switching.
TEM-1 β-lactamase — Cryptic allosteric site
Major antibiotic resistance enzyme. Allosteric inhibition offers a path to overcome resistance mutations in the active site.
PTP1B — Allosteric cryptic site
Diabetes and obesity target. The catalytic site is notoriously undruggable; this allosteric pocket provides an alternative entry point.
Calmodulin — Cryptic hydrophobic pocket
Universal calcium signaling mediator. Hidden hydrophobic cleft only exposed upon conformational change, enabling selective modulation.
HIV-1 protease — Flap dynamics cryptic site
Critical antiviral target. Flap region dynamics reveal transient pockets for next-generation protease inhibitors that evade resistance.
See It in Action
Submit your protein sequence and discover binding sites that no static method can find.
Try Free Analysis