Designing a Selective IL‐2 Binder with RFDiffusion, ProteinMPNN, and Boltz‐1

Read our de novo cytokine receptor binder tutorial here →

Dear all,

Interleukin‑2 (IL‑2) was one of the first immunotherapies to demonstrate that immune activity alone could shrink tumors. But the cytokine therapeutic was poorly understood in the late 1900s when it was discovered, and it’s repurposing as a therapeutic came with trade-offs: severe toxicities, immune suppression, and a narrow therapeutic window. These problems stem from how IL‑2 interacts with its cognate receptor, the IL2R αβγ complex. IL2R contains two functionally distinct receptor chains, the βγ-heterodimer and then IL2-Rα receptor. In recent years, it is understood that IL2’s interaction with the βγ-heterodimer mediates most of IL2’s therapeutic effects, while the IL2 → IL2-Rα axis is responsible for toxicity and anti-tumor immunosuppression.

Enter our question for this tutorial: Can de novo protein design enable the design of a more selective IL2R agonist, emphasizing therapeutic impact while minimizing the downsides of endogenous IL2 signaling?

To explore this design challenge, we use Superbio’s protein engineering toolkit to develop candidate IL‑2 variants with selective βγ binding. RFDiffusion generates backbone scaffolds using key hotspot residues from the IL‑2/βγ interface. ProteinMPNN translates those backbones into foldable amino acid sequences. Finally, Boltz‑1 ranks each design by predicted interaction strength and receptor selectivity - prioritizing those with minimal α-chain engagement. The result is a no-code, open-source pipeline for generating and scoring selective cytokine variants in under an hour.

Inside the tutorial [link here]
*Examine target protein PDB, defining target regions for protein design
* 3D backbone generation with RFDiffusion → ProteinMPNN sequence‑optimization
* Rank designs with Boltz‑1 interface scores
* Filter candidate constructs - all in under an hour on Superbio

Want a different model or use case next? Hit reply and let us know.
Until next time 💗,
The Superbio Team