Precomputed demo

AlphaGene Protein Design Studio

From target structure to designed protein candidates.

Explore a prepared target-to-design run. This page replays previously computed RFdiffusion and ProteinMPNN outputs; it does not start a live GPU job.

  1. 01Define targetKv1.3 channel interface
  2. 02Generate backboneRFdiffusion candidate set
  3. 03Design sequenceProteinMPNN with fixed motif
  4. 04Review candidateStructure and method checks
Experimental reference structures

Classic Binding Atlas

See how established ligands and inhibitors engage their targets.

Explore nine landmark binding stories across oncology, metabolism, immunology, and infectious disease. Each view replays a public experimental structure from the Protein Data Bank—not a newly generated AlphaGene result.

Immuno-oncology

PD-1 checkpoint blockade

Compare the natural checkpoint interface with clinically established antibody complexes.

PDB 4ZQK
Experimental complex Drag to rotate · pinch or scroll to zoom · right-drag to move
Loading experimental structure Reading coordinates from the local case library.
0%

Three-step explanatory view · not a molecular-dynamics trajectory.

Target Binding partner Interface residues
Binding logic

Natural checkpoint recognition

Experimental record
Method
Resolution
Displayed chains
Interpretation boundary

What this structure does—and does not—show

Reference-library boundary

These are public experimental structures used to explain molecular recognition. Static coordinates do not establish affinity, selectivity, efficacy, safety, or real-time binding dynamics. Antibody entries show Fab fragments where indicated; receptor entries may show only experimental domains or ectodomains. View structure provenance.

Interactive workspace

Replay a real prepared binder-design example.

Prepared run · no GPU cost

Interactive 3D complex

AG-KV13-001

Stage 04 / 04 Review candidate Illustrative transition · not RFdiffusion timesteps
Loading prepared structure Parsing target and binder coordinates in your browser.
Drag to rotate · scroll to zoom · right-drag to move
Prepared result ready 100%

Prepared candidate set

Compare backbone and sequence outputs.

Choose a card to load its complex, sequence, metrics, and downloads.

ProteinMPNN sequence

AG-KV13-001

Prepared output

Sequence length
Backbone Rg
ProteinMPNN score
Backbone assessment
Download PDB

Structure-consistency review

Evidence available in this run

Backbone only
  • Included
    RFdiffusion backbone complexTarget chain B and designed binder chain A
  • Included
    ProteinMPNN sequence designFixed Lys–Tyr motif and recorded model score
  • Not run
    Independent fold-back predictionNo AlphaFold, ColabFold, or ESMFold score is claimed
  • Not run
    Experimental binding validationNo affinity, specificity, safety, or efficacy data

Computational structure-consistency checks do not establish binding affinity, specificity, stability, safety, efficacy, or experimental validation.

Methods and provenance

Every displayed artifact has a traceable role.

01

Public target system

A 7SSZ-derived Kv1.3 channel/toxin system supplies the prepared target geometry and interface context.

02

RFdiffusion backbones

Three real outputs from the official RosettaCommons binder-design tutorial form the candidate set.

03

ProteinMPNN sequences

Sequences were generated for each binder backbone with the tutorial's Lys–Tyr motif held fixed.

04

Optional live validation

A private GPU backend can add fold-back prediction, complex scoring, queues, and project persistence.

From demo to live platform

Connect a private GPU worker when you are ready to calculate new designs.

The current experience is a zero-GPU-cost prepared demonstration. A live deployment can keep the same interface while adding customer accounts, private uploads, job queues, and project history.

Discuss a private design run