TNF-α
Tumor necrosis factor
Design a conditional TNF-α binder
TNF-α signals as a trimer, with three receptor-binding sites where its subunits meet. Antibody drugs such as adalimumab and infliximab, used in rheumatoid arthritis and Crohn's disease, neutralise it by blocking those sites. This challenge adds two requirements to plain binding. The design must recognise mouse TNF-α as well as human, and it must hold human TNF-α at neutral pH but let go at acidic pH.
Submissions closed Sunday, October 18, 23:59 AoE. Selection begins shortly.
Human TNF-α trimer, apo form · PDB 1TNF
Target information
Design against the soluble, trimeric form of TNF-α. We recommend targeting a functional epitope such as the receptor-binding site at the interface between adjacent protomers.
Human TNF-α
Soluble trimer- UniProt reference
- P01375-1 ↗
- Target residues · UniProt
- 77–233 / 157 aa
- Assay construct
- TNA-H4211 · tag-free ↗
- Recommended epitope
- Receptor-binding site between two protomers
- Structure reference
- 1TNF · Chains A–C ↗
Amino acid sequence
VRSSSRTPSDKPVAHVVANPQAEGQLQWLNRRANALLANGVELRDNQLVVPSEGLYLIYSQVLFKGQGCPSTHVLLTHTISRIAVSYQTKVNLLSAIKSPCQRETPEGAEAKPWYEPIYLGGVFQLEKGDRLSAEINRPDYLDFAESGQVYFGIIALBoth constructs are supplied as trimers (SEC-MALS: about 45–62 kDa human, 50–65 kDa mouse), so each target presents three copies of the epitope. The two soluble domains are 79% identical (124 of 156 aligned residues).
The mouse sequence leaves out the C-terminal polyhistidine tag carried by the assay construct.
The targets' purification tags and your design's C-terminal Twin-Strep capture tag are off limits: binding to a tag does not count.
Three objectives
This challenge consists of 3 main objectives, and each will be accounted for when selecting the best designs.
01
Binding affinity
Design a binder against human TNF-α. We recommend targeting a functional epitope such as the receptor-binding site at the interface between adjacent protomers, which is also targeted by the therapeutic antibodies adalimumab and infliximab.
We'll measure its binding affinity using the soluble trimeric form of human TNF-α.
02
Mouse cross-reactivity
In the next objective, you should design a binder that recognises both human and mouse TNF-α. On top of the validation from Objective 1, we will test binding against mouse TNF-α at pH 7.4.
Binding both species would make it easier to study the same design in mouse models before moving to human studies, without developing a separate mouse-specific binder.
How designs are ranked
Every tested design is measured in the same four conditions: human TNF-α at pH 7.4 and at pH 6.0, and mouse TNF-α at pH 7.4, with the design immobilised and the target in solution. A design counts as binding in a condition when its sensorgram gives a fittable curve and a KD, or, where the curve cannot be fitted, when the association shift clears 300% over the negative control. Otherwise we call it no detectable binding.
- 1
pH-selective binding
A measurable KD against human TNF-α at pH 7.4 and no detectable binding at pH 6.0. Designs that bind at both pH values rank after these, by the ratio of their KD at pH 6.0 to their KD at pH 7.4. Designs that rely on a conformational change or a designed pH mechanism are not penalised when we select designs for testing; the measurement decides.
- 2
Mouse cross-reactivity
Binds mouse TNF-α as well as human at pH 7.4, ranked by how close the mouse-to-human KD ratio is to 1, among designs whose human KD is below 1 µM.
- 3
Affinity
Lowest KD against human TNF-α at pH 7.4. Binding at a functional epitope counts in its favour: name the epitope you target and link your evidence, or we read it from a predicted complex. The recommended epitope is a hint, not a rule: a design that binds elsewhere is measured and ranked all the same.
The winner has to clear all three steps, with ties settled in this order, and each step also awards an honourable mention to the best design that did not win.
Which designs are tested is a separate step from who wins. Track 1 teams get their ranked designs tested; for Tracks 2 and 3, Claude selects from all submissions with a prompt we publish in full.
Timeline
Oct 5 – 18, 2026
Submissions open
Design, review, and submit. Submissions close Sunday, October 18 at 23:59 Anywhere on Earth (AoE).
- 2nowSee other challenges
Submissions closed
The window is closed. No further designs are accepted for this challenge.
- 3
Selection in progress
Anthropic and Adaptyv select designs for synthesis using a Claude-based workflow, published after the competition.
- 4
Wet-lab validation in progress
Submission requirements
- Length
- 10–250 amino acids
- Format
- Single chain protein, nanobody, scFv, or Fab
- Designs per participant
- Up to 40 for Track 1, up to 20 for Tracks 2 and 3
- Design
- De novo designs only