Targeting challenging protein from integrin family

Utilizing Receptor.AI’s pocket detection workflow to target the ‘undruggable’ integrin family

~8M

compounds
screened

~12%

hit rate with 97 candidates

3

binding pockets
targeted

<10 μM

IC50 of 32 best hit compounds

0.35 μM

dissociation constant
of the top compound

*Project workflow

01/ Background

  • Target is a heterodimeric receptor from the integrin family.
  • It is a promising target in immunological disorders and cancers.
  • There were no binding sites for small molecules known.
  • Also, no potent small molecule binders were known.
  • Target is often deemed undruggable by small molecules.
  • The goal is to design a strong binder to further optimize it into
PPI disruptor.

02/ Methodology

  • Conformational changes predicted with crystal structures using AI-driven MD simulations.
  • Binding pockets identified within dynamic conformations by proprietary AI model.
  • Virtual screening of 8M stock library.
  • AI-docking with target-specific rescoring using ArtiDock.
  • In vitro ligand displacement assay performed to validate hit candidates’ ability to compete with a known ligand for binding.
  • SPR analysis performed to measure binding kinetics and affinity (dissociation constant — Kd).

03/ Results

  • Identified 97 hits out of ~1,200 hit candidates, resulting in 
12% hit rate.
  • Identified 4 binding pockets, including one cryptic.
  • Successfully targeted 3 pockets - Pocket I (41 hits), Pocket II
(40 hits), and Pocket III (16 hits).
  • 32 hit compounds with IC50 < 10 μM.
  • The top hit shows the highest affinity (Kd = 0.35 μM).
*Pocket I visualization
*Pocket II visualization
*Pocket III visualization