STRC h26 Phase 1d AF3 Final 2026-04-27

The last missing Phase 1d AlphaFold Server result, engineered_homodimer_A1078C, was retrieved from AFS fold 27162e77fc95333e and downloaded to models/fold_engineered_homodimer_A1078C_afs.zip.

Result

Gate G1 requires ipTM >= 0.50 and pTM >= 0.60. A1078C fails decisively across all five AF3 models.

Modelranking_scoreipTMpTMG1
00.340.250.39FAIL
10.320.240.37FAIL
20.310.240.40FAIL
30.310.230.39FAIL
40.300.230.39FAIL

Phase 1d final set:

MutantipTMpTMG1AFS fold
A1078C0.250.39FAIL/fold/27162e77fc95333e
S1080C0.330.41FAIL/fold/73b2ae8798b7a3fe
S1579C0.230.39FAIL/fold/69aa00e31b6c2c6f
A1078W negative control0.230.37FAIL/fold/5403b988473e6dca

Interpretation

The disulfide-engineering rescue path is closed. Phase 1 already falsified the ARM aromatic/arginine-repulsion idea. Phase 1c found plausible cysteine geometry, but Phase 1d shows that the best remaining cysteine candidates do not produce an AF3-supported homodimer complex.

No G2 contact parser, G4 energy job, G5 MD, or G9 h03 drop-in test is justified because no candidate clears the upstream AF3 confidence gate.

Ranking delta - 2026-04-27

h26 before: tier=B, mech=2, deliv=5, misha_fit=4, next=Retrieve/finish Phase 1d A1078C AF3; if FAIL, demote h26 to C; if PASS, parse contacts then score G4/G5 as h03 drop-in candidate.

h26 after: tier=C, mech=2, deliv=5, misha_fit=4, next=Paused: Phase 1d A1078C/S1080C/S1579C/A1078W all fail AF3 G1; no further compute unless a new scaffold/interface appears.

Evidence: A1078C final AF3 best model is ipTM 0.25 / pTM 0.39 against a 0.50 / 0.60 G1 gate; all four Phase 1d candidates fail. The only fast B-to-S route for h26 is gone, so h26 leaves B-tier.

Connections