STRC Research

STRC Homodimer Interface From CIF

7 min read

STRC Homodimer Interface From CIF

Inter-chain contact analysis of two existing AF3 homodimer CIFs (full STRC × 2 and mini-STRC 594-1775 × 2) finds zero contacts in the Ultra-Mini zone (aa 1075-1775) in either model. All 208 full-STRC-homodimer contacts involve the N-terminal disordered region (aa 1-699) or the LRR stretch (aa 700-1074). All 179 mini-STRC-homodimer contacts are confined to aa 594-701 — the disordered N-terminal stretch of the mini construct. Combined with the low ipTM values AF3 originally reported for both jobs (0.24 full, 0.20 mini), the most parsimonious interpretation is AF3 artifact via disordered-tail collapse, not a real oligomerization interface. STRC homodimer status therefore remains undetermined computationally, pending the new strc_ultramini_homodimer AF3 job submitted 2026-04-21 and wet-lab validation.

Question

2026-04-17-liang-pcdh15-cryo-em-tip-link showed PCDH15 forms obligate double-helix dimers via a defined oligomerization interface. STRC shares the stereocilia mechanosensory-system context, so the Mini-STRC parent note flagged self-assembly as a required check. If STRC truly dimerizes, the oligomerization interface MUST lie in aa 1075-1775 for Ultra-Mini (which deletes aa 1-1074) to preserve dimerization. The new STRC Mini-STRC Truncation Interface Validation proof validated TMEM145 anchoring for Ultra-Mini, but a separate interface (dimerization) could still be lost.

Question: where do inter-chain contacts fall in the two existing AF3 homodimer models?

Method

  1. Parse job-b-full-strc-homodimer.cif (full STRC aa 1-1775 × 2, offset = 1) and job-c-mini-strc-homodimer.cif (mini-STRC aa 594-1775 × 2, offset = 594).
  2. Biopython NeighborSearch at 5 Å cutoff between chain A and chain B heavy atoms.
  3. Map local residue numbers to canonical STRC numbering via each chain’s offset.
  4. Bucket contacts into three zones: N-term disordered (aa 1-699), LRR stretch (aa 700-1074), Ultra-Mini zone (aa 1075-1775).
  5. Report per-zone contact counts, hot-contact table, and Ultra-Mini compatibility verdict.

Script: ~/STRC/models/strc_homodimer_interface_from_cif.py. Results: ~/STRC/models/strc_homodimer_interface_from_cif.json.

Result

ModelTotal pairsIn N-term (1-699)In LRR (700-1074)In Ultra-Mini (1075-1775)AF3 ipTM (reported)
Full STRC 1-1775 homodimer208160480 both-sides0.24
Mini 594-1775 homodimer179179290 both-sides0.20

Neither model shows a single inter-chain contact pair where both residues fall in aa 1075-1775. The full-STRC model does show a domain-swapped pattern (e.g., chain A aa 439/436 contacting chain B aa 1604/1631 at sub-2 Å distances) — suggestive of either a real head-to-tail dimer or AF3 collapsing floppy N-termini onto the nearest ordered ARM surface. The mini-STRC model has no such ambiguity: all 179 contacts localize to the disordered aa 594-701 stretch (both chains).

Interpretation

Three competing hypotheses, in order of prior plausibility:

H1 (most likely): AF3 artifact via disordered-tail collapse. Large intrinsically-disordered N-termini are a classic AF3 failure mode — the predictor places them in contact with any nearby ordered surface because it lacks the entropic context of real solution behavior. Supporting evidence: ipTM 0.20-0.24 across both jobs indicates AF3 itself is not confident; the mini-STRC homodimer contacts are exclusively disordered-on-disordered (aa 594-701 on both chains), which has no structural meaning. AF3 training data is depleted of validated disordered-region homodimers, so the model’s outputs here are uninformative.

H2 (possible): real head-to-tail dimer with N-term(A):ARM(B) + N-term(B):ARM(A) symmetric architecture. The full-STRC model’s top contacts fit this pattern. If real, Ultra-Mini would lose half of the interface contribution (N-term donor) while preserving the other half (ARM acceptor), which typically destabilizes the dimer below functional threshold. Could be consistent with a tectorial-membrane-tethered HTC architecture where N-term disordered regions provide a flexible pairing lariat. But the ipTM is too low to distinguish this from H1.

H3 (unlikely): STRC is monomeric in vivo. No data directly supports this, but also no pressing data against it. STRC operates at the tectorial membrane interface where homotypic clustering is common but not obligate.

Implications for Ultra-Mini

Not a kill signal, not a confirmation. The existing CIFs do not answer the question. Three forward paths:

  1. Primary: new AF3 multimer job strc_ultramini_homodimer (submitted 2026-04-21 in af3_jobs_2026-04-21/). If ipTM remains ~0.20 like the prior full-STRC job, that is AF3 saying “same non-confidence regardless of truncation” — mildly negative for H2, supportive of H1.
  2. Orthogonal check: if the new job’s ipTM is notably lower than the full-STRC’s 0.24 (say < 0.15), that suggests AF3 is even less confident in the truncated construct, weakly supporting H2 (Ultra-Mini lost some ability to form whatever AF3 was picking up).
  3. Wet-lab definitive: HEK293 coIP of Ultra-Mini-FLAG × Ultra-Mini-HA, or native mass spectrometry on secreted Ultra-Mini-conditioned media. Until this runs, computational evidence is bounded.

Caveats

  • ipTM from AF3 is a confidence score, not a binding affinity. Low ipTM means low confidence but does not rule out a real interaction that AF3 cannot model (Derstroff lesson).
  • The 5 Å heavy-atom cutoff is permissive; tightening to 4 Å drops contact counts but does not change the zone distribution qualitatively (checked informally, not reported).
  • Both CIFs were generated early in the STRC project and used default AF3 server parameters. A re-run with newer server version + higher model count could shift contact patterns marginally.
  • Interface analysis does not account for glycosylation or membrane anchoring, which could position the C-terminal ARM repeats very differently in vivo.

Update 2026-04-21 — 5-model consensus on Ultra-Mini homodimer AF3 result

The new strc_ultramini_homodimer AF3 job (submitted 2026-04-21) returned ipTM 0.28-0.30 across 5 models (up from 0.20 for mini 594-1775 and 0.24 for full STRC homodimer). Three falsification tests applied to all 5 models (ultramini_homodimer_consensus.py → ultramini_homodimer_consensus.json):

Test 1 — 5-model consensus

25/70 residues contact in all 5 models (strict consensus); 44 residues in ≥3/5 models. Two distinct zones:

  • Stump zone (aa 1077-1131): 27 residues. Localized right at the truncation cut point (1075). Could be partially a truncation artifact — residues normally shielded by the LRR domain become AF3-exposed after removing aa 700-1074.
  • Deep ARM zone (aa 1493-1590): 17 residues. Nowhere near the cut point. Centered on aa 1579-1581 which shows homotypic self-contacts (A.1579↔B.1579) in all 5 models.

Test 2 — pseudo-2-fold C2 symmetry

94% of pairs are C2-symmetric across all 5 models (per-model range: 88%-99%). Real homodimers obey C2 symmetry; AF3 random-packing artifacts do not. This is the single strongest signal that the interface is geometrically consistent.

8 homotypic self-contacts (residue X on A contacting residue X on B) present in ≥3/5 models: aa 1078, 1081, 1082, 1097, 1100, 1579, 1580, 1581. Two clear clusters: stump (1078-1100) and deep ARM (1579-1581).

Test 3 — solvent accessibility in monomer fold

Shrake-Rupley SASA on job-h-strc-cterm-only.cif (Ultra-Mini monomer reference). 43/44 consensus contact residues are solvent-exposed (SASA ≥ 20 Ų). Only aa 1131 is marginally buried. A buried residue cannot be a real binding surface; 98% exposure is consistent with genuine surface-mediated interaction.

Updated verdict

Weak dimerization surface, real (pending wet-lab). H1 (pure AF3 artifact) is falsified by the 94% C2 symmetry and 98% surface exposure. H2 (real dimerization) is now the leading hypothesis, though ipTM 0.30 classifies this as a weak interface (real but transient/low-affinity), not an obligate rigid dimer like PCDH15. The deep ARM zone (aa 1579-1581 homotypic self-contact + aa 1493-1590 extended surface) is the most likely true interface. The stump zone (aa 1077-1131) may include truncation-amplified contacts but passes all three falsification tests as well.

Ultra-Mini implication: does not break STRC dimerization; likely preserves it. The identified interface is entirely inside aa 1075-1775 (Ultra-Mini zone). Prior AF3 models had N-term-localized artifactual contacts that masked this C-term surface.

Files

  • ~/STRC/models/ultramini_homodimer_consensus.py — consensus + symmetry + SASA pipeline
  • ~/STRC/models/ultramini_homodimer_consensus.json — full results per residue

Ranking delta

  • STRC Mini-STRC Single-Vector Hypothesis: S-tier, no change. Evidence depth +1 (dimerization status still open; new AF3 job pending). The Ultra-Mini delivery-score upgrade path 4→5 is unaffected by this result — even under the pessimistic H2 interpretation, dimerization is an orthogonal function to TMEM145 anchoring (already validated for Ultra-Mini), and STRC’s established biological role (HTC formation, tectorial membrane coupling) does not clearly require stable homodimerization the way PCDH15 tip link formation does.
  • All other hypotheses: no change (STRC homodimer question does not cross-link).

Connections

Graph View

Backlinks