PD-L1 Target Deep Dive
Immune Checkpoint Inhibitor Design
This lesson uses the course’s canonical PD-L1 record: 4ZQK, chain A. Do not substitute the older 5O45 exercise record without re-deriving chain IDs and hotspots. The machine-readable source of truth is data/targets.yml.
Biological Context
Programmed Death-Ligand 1 (PD-L1) is a critical protein expressed on the surface of many cells, including cancer cells. Under normal conditions, it acts as a “brake” on the immune system. When PD-L1 binds to the PD-1 receptor on T-cells, it signals the T-cell to become inactive.
Why it matters: Cancer cells often hijack this mechanism by over-expressing PD-L1. This allows them to “trick” the immune system into ignoring the tumor. Therapeutic antibodies against PD-L1 (atezolizumab, durvalumab) have transformed oncology, but they are large, expensive to manufacture, and can trigger autoimmune side effects — a smaller, more controllable de novo binder would be a meaningful advance.
The Goal: Design a protein binder that competes with PD-1 for the same interface on PD-L1. Blocking that interaction releases the “brake,” allowing T-cells to recognize and attack the cancer.
Interactive Structure
Explore the native interaction between PD-L1 (the target) and PD-1 (the natural binder) below.
- Target (PD-L1): Chain A — the surface we want to bind to.
- Binder (PD-1): Chain B — the natural partner we want to compete with.
Non-interactive alternative: The target-specification table below describes the relevant chains and interface residues. You can also open the 4ZQK structure record or download its PDB coordinates. Manipulating the 3D viewer is optional.
- Rotate: Left-click and drag
- Zoom: Scroll wheel
- Pan: Right-click (or Ctrl+Left-click) and drag
Design Mission
Your objective is to create a de novo protein binder that binds to the same interface on PD-L1 that PD-1 currently occupies.
Target Specifications
| Feature | Detail |
|---|---|
| Target Name | PD-L1 (Programmed Death-Ligand 1) |
| PDB ID | 4ZQK |
| Target Chain | Chain A (PD-L1 IgV domain, UniProt residues 18–132) |
| Partner (to compete with) | Chain B (PD-1) |
| Entry caveat | Target chain A is unmutated; partner PD-1 chain B carries C93S |
| PDB-derived interface contacts (≤5 Å) | A18, F19, T20, V23, D26, I54, Y56, E58, N63, Q66, V76, R113, M115, S117, G119–I126 |
| Candidate steering set | A56,A58,A113,A122,A123 (Y56, E58, R113, D122, Y123), selected from dense contacts on the PD-1-facing surface |
These residues are every PD-L1 position with any heavy atom within 5 Å of PD-1 in deposited 4ZQK. That makes them geometric interface contacts, not measured energetic hotspots. The smaller course set is a reproducible way to steer generation toward the PD-1-facing surface; it does not establish which contacts contribute most to affinity.
Strategy Tips
- Download PDB
4ZQK. - Clean the structure: Keep Chain A (PD-L1) as the target. Remove Chain B (PD-1) and any water molecules.
- Define steering residues: When running RFdiffusion or BindCraft, begin with
A56,A58,A113,A122,A123. This biases diffusion toward the PD-1-competing face rather than a random patch. - Validate with AlphaFold2/Chai-1: After designing a binder sequence, predict the complex and check that the proposed interface covers the intended PD-1-facing region.
Reference
- Zak, K.M. et al. (2015). Structure of the complex of human programmed death 1, PD-1, and its ligand PD-L1. Structure 23, 2341–2348. doi:10.1016/j.str.2015.09.010