STRC AF3 Truncation Boundary Sweep
Purpose
Refine the optimal N-terminal truncation point for mini-STRC. Previous work established 700 as cut point based on pLDDT dip at 685-694. These jobs test ±50 residues around it.
Results (2026-03-19)
All jobs: 5 seeds each, single chain, protein-only.
| Cut point | Residues | aa | CDS (bp) | pTM | Ranking | Disorder | Headroom |
|---|---|---|---|---|---|---|---|
| 650 | 650-1775 | 1126 | 3378 | 0.84 | 0.88 | 7% | 1322 bp |
| 680 | 680-1775 | 1096 | 3288 | 0.84 | 0.87 | 6% | 1412 bp |
| 700 | 700-1775 | 1076 | 3228 | 0.86 | 0.88 | 4% | 1472 bp |
| 720 | 720-1775 | 1056 | 3168 | 0.86 | 0.89 | 4-5% | 1532 bp |
Reference (from previous jobs):
| Construct | pTM | Ranking | Disorder |
|---|---|---|---|
| Full STRC (1-1775) | 0.63 | — | 16% |
| Mini-STRC (616-1775) | 0.81 | 0.85 | 7% |
| Shorter mini (700-1775, prev) | 0.86 | 0.88 | 4% |
Key Findings
- 700 and 720 are both optimal. pTM 0.86, disorder 4-5%. 720 has marginally better ranking (0.89 vs 0.88).
- 650 and 680 are worse. Including the transition zone (650-699) adds disorder without structural benefit.
- pLDDT dip at 685-694 is a real domain boundary. Cutting before it (650, 680) captures disordered residues.
- 720 could be even better. 20 fewer residues, slightly better ranking, more AAV headroom.
- All 4 jobs show rock-solid seed consistency (all 5 seeds within 0.01 pTM).
Decision
Keep 700 as primary. It’s at the exact recovery point of the pLDDT dip and includes the full LRR domain start. 720 is a valid alternative if we need more headroom.
IgK-SP + mini-STRC (700-1775)
Separate job to test if prepending signal peptide affects the fold:
| Metric | Without SP (prev) | With IgK-SP | Delta |
|---|---|---|---|
| pTM | 0.86 | 0.85 | -0.01 |
| Ranking | 0.88 | 0.88 | 0 |
| Disorder | 4% | 5% | +1% |
SP does not significantly affect the fold. The 0.01 pTM drop is within noise. IgK is cleaved co-translationally anyway, so the mature protein won’t contain it.
Codon-Optimized Job (INVALID)
fold_igk_sp_shorter_mini_strc_codon_optimized — pTM 0.23. DNA sequence (ATGC…) was submitted as proteinChain. AF3 interpreted nucleotide letters as amino acids. Result is garbage, discard entirely.
Files
- CIF models:
~/Sites/site-strc-egor-lol/public/models/job-truncation-{650,680,720}.cif,job-igk-sp-mini-strc.cif - Raw data:
~/DeepResearch/strc/af3-results/job-truncation-{650,680,720}/,job-igk-sp-mini-strc/ - Invalid:
~/DeepResearch/strc/af3-results/job-igk-sp-codon-opt/(DNA submitted as protein)
Connections
[supports]STRC pLDDT Profile and Cut Point Analysis — confirms 700 is optimal boundary[supports]STRC Mini-STRC Single-Vector Hypothesis — refines the candidate[supports]STRC Signal Peptide Validation — SP doesn’t break fold- STRC AAV Vector Design — cut point confirmed
[about]Misha