4-AP Reframed: Selective Axonal Kv1.1/Kv1.2 Blocker, NOT Multi-Target Recovery Agent
Paradigm correction (the second one in a week)
After three new DiffDock batches and an orphan-trajectory analysis, the 4-AP story is clearer — and smaller — than either of our prior framings:
- April 2 framing (wrong): "4-AP is a multi-target recovery agent that hits CORO1C, NCALD, SMN2, SMN1, UBA1, and regenerates motor neurons."
- April 6 framing (still wrong): "4-AP is a complementary recovery agent that supports BDNF/TrkB/STAT3/PTEN/mTOR signaling."
- April 10 framing (correct): 4-AP is a selective axonal Kv1.1/Kv1.2 blocker. That is the Ampyra mechanism. There is no structural support for the regeneration story.
What we ran
Three DiffDock v2.2 batches on AWS NIM:
Batch 1 — Regeneration target panel (4-AP, 5 targets)
BDNF (1BND), TrkB (4ASZ), STAT3 (1BG1), PTEN (1D5R), mTOR FRB (4DRI). All five rank-1 confidences between -0.16 and -0.38 (BDNF reaches -2.96 only because BDNF is a small homodimer with broad pockets that pick up almost anything). None of these targets is a real 4-AP binding partner. The regeneration story is structurally unsupported.
Batch 2 — Kv1 family selectivity (4-AP, 3 targets)
Kv1.1 (6EBK), Kv1.3 (3OC3), Kv1.5 (7SIT) — to test whether 4-AP discriminates axonal Kv1.1/Kv1.2 from cardiac Kv1.5 and lymphoid Kv1.3.
- Kv1.1: -0.05
- Kv1.3: -0.78
- Kv1.5: -0.10
Combined with the April 2 Kv1.2 result, 4-AP shows the strongest engagement at the axonal Kv1.1/Kv1.2 pair with modest but real selectivity over Kv1.3 / Kv1.5. This is consistent with the established Ampyra mechanism.
Batch 3 — Head-to-head vs 3,4-DAP / Firdapse (13 targets)
The full panel re-run with 3,4-diaminopyridine (Nc1ccnc(N)c1).
| Target | 4-AP | 3,4-DAP | Better |
|---|---|---|---|
| Kv1.1 | -0.05 | -0.17 | 4-AP |
| Kv1.2 | (April 2 +) | -0.52 | 4-AP |
| Kv1.3 | -0.78 | -1.17 | 4-AP |
| Kv1.5 | -0.10 | -0.30 | 4-AP |
| TrkB | -0.19 | -0.69 | 4-AP |
4-AP outperforms 3,4-DAP at every Kv1 subtype. This matches clinical practice: 4-AP/Ampyra is the preferred MS walking drug despite 3,4-DAP being more potent in some Lambert-Eaton contexts.
Bonus from the same day: orphan trajectory analysis rediscovered SMN2 binding
Independently, an analysis of 50 orphan MD trajectories from the cluster found that 4-AP also weakly engages a novel SMN2 Tudor-domain pocket (PRO268, SER271, TYR657 region) — a finding that was hidden by a topology atom-count artifact in the original analysis. The same pocket is hit by Riluzole. So the final story is:
"Selective axonal Kv1.1/Kv1.2 blocker WITH a secondary druggable SMN2 pocket binding event shared with Riluzole."
This is richer than "pure Kv blocker" and structurally honest, which the April 2 "multi-target recovery agent" framing was not.
Why publish a correction
Negative results matter. Two earlier framings of 4-AP were too generous; this one is calibrated to the structural data. Publishing the reframe is the same scientific-integrity discipline as publishing the negative Fasudil scaffold-hop result.
Where the data lives
- GitHub (3 batch summaries + README): https://github.com/Bryzant-Labs/sma-research/tree/main/campaigns/4-AP/2026-04-10_updates/diffdock_extensions
- Dropbox open_data (all SDF poses):
Dropbox/SMA/open_data/4ap_extensions_2026-04-10/ - Orphan trajectory analysis: https://github.com/Bryzant-Labs/sma-research/blob/main/findings/2026-04-10/ORPHAN_TRAJECTORY_ANALYSIS.md
CC-BY-4.0. Compute cost: ~$3.