FOLD №58 — FAILED

Epitalon N-to-C Hybrid: AEDG-AEEAc-GRKKRRQRRR-NH2

TLDR

This fold was FAILED due to a clean technical infrastructure failure: Boltz-2 returned a 400 error and produced no output files; Chai-1 was disabled. No structure, confidence scores, or binding metrics were generated. This is a tool-level failure, not a biological verdict — the chimeric hypothesis was never evaluated.

What we tried

Every prior Epitalon fold in this laboratory — D-Ala substitution (Fold №6), C-terminal amidation (Fold №21), head-to-tail cyclization (Fold №26), pyroglutamate capping (Fold №35), and palmitoyl-γGlu-Lys lipidation (Fold №44) — returned pLDDT 0.34, the structural sub-resolution floor for a 4-mer. These were not biological verdicts; they were consistent demonstrations that AlphaFold-family tools cannot resolve backbone geometry for sequences this short.

FOLD №58 attempted to escape this floor entirely by embedding Epitalon (AEDG) in a chimeric construct: free N-terminus of AEDG, followed by a PEG2 spacer (AEEAc, 8-amino-3,6-dioxaoctanoic acid), followed by the cell-penetrating peptide TAT(48-57) (GRKKRRQRRR), C-terminally amidated. The resulting ~14-residue hybrid was hypothesized to (a) provide sufficient sequence context for meaningful structural prediction, and (b) deliver the AEDG pharmacophore to intracellular/nuclear TERT via the extensively validated TAT macropinocytosis/translocation pathway.

This was also the first Epitalon fold in the lab to address the delivery problem rather than protease stability alone — a strategic departure motivated by literature evidence (PMID:40908429; PMID:39788414) that TERT is a nuclear target and that Epitalon's membrane permeability is entirely uncharacterised.

Why it was discarded

The Boltz-2 prediction subprocess returned a 400 error and wrote no output files. Chai-1 was not available as a fallback. The failure is technical and infrastructure-level: the predictor did not run to completion. No pLDDT, pTM, ipTM, or binding affinity values exist for this fold.

The heuristic sequence-based profile (aggregation propensity 0.0, stability score 0.307, no BBB penetration, moderate-to-long half-life estimate) was generated from sequence properties alone and cannot substitute for a folded structural model or confidence metrics.

Notably, the PEG2 linker (AEEAc) introduces non-standard chemistry that some AlphaFold-family implementations handle inconsistently; this may have contributed to the subprocess error, though the precise cause was not diagnosed.

What this doesn't mean

FAILED does not mean disproved. The chimeric hypothesis — that fusing AEDG to TAT(48-57) via a PEG2 spacer would enable intracellular delivery to nuclear TERT while providing sufficient sequence context for confident structural prediction — was never evaluated. The predictor did not produce output; it did not model the construct and find it non-binding or disordered. The biological rationale remains intact: TAT-mediated nuclear delivery is extensively validated, PEG2 linker chemistry is standard, and the literature supports nuclear TERT as a relevant target for Epitalon-class peptides. This fold represents a genuine strategic and chemical advance over all previous Epitalon folds in this lab, and its failure is entirely an artifact of computational infrastructure, not of the underlying science.

There are legitimate biological cautions that would need to be addressed in any future evaluation — including the possibility that the AEDG acidic patch is intramolecularly sequestered by TAT's cationic residues across the PEG2 spacer, that AEDG acts via histone chromatin remodelling upstream of TERT rather than direct TERT engagement (PMID:32019204), and that the primary mechanistic paper has a correction notice (PMID:41240216) — but none of these constitutes a structural disproof of the construct.

What would answer the question

• Re-run with stable Boltz-2 infrastructure or RoseTTAFold2-AA, which has demonstrated capacity to handle non-standard linker chemistries (AEEAc) as CCD components; alternatively, model the TAT segment alone first to confirm the predictor accepts the sequence, then resubmit the full chimera.
• Chai-1 ensemble prediction (when available): the polycationic TAT helix and the flexible PEG2 tether represent a context where multi-model ensemble scoring would add meaningful confidence beyond a single-run pLDDT.
• Molecular dynamics (MD) simulation of the assembled chimera to assess whether the PEG2 spacer provides sufficient conformational freedom to prevent intramolecular AEDG–TAT electrostatic sequestration — this is a critical prior question before any wet-lab synthesis.
• Cellular NMR or fluorescence co-localisation of a labelled AEDG-PEG2-TAT conjugate in hTERT-expressing fibroblasts (e.g., IMR-90 cells used in PMID:40908429) to directly test nuclear delivery and TERT co-localisation — the most direct experimental adjudication of the hypothesis.
• Surface plasmon resonance (SPR) or ITC with recombinant hTERT catalytic domain to test whether AEDG alone, or the chimera, binds TERT directly — this would resolve the outstanding question of whether TERT engagement is direct or indirect (histone-mediated).

Raw metrics

All heuristic values are sequence-based estimates only — not structural predictions or wet-lab measurements.

In silico prediction only. No wet-lab validation performed. This is research, not medical advice.