Selank (TKPRPGP) is a synthetic heptapeptide analog of tuftsin developed at the Institute of Molecular Genetics (Russian Academy of Sciences), combining the immunomodulatory TKPR tetrapeptide core with a C-terminal PGP stabilizing extension. The TKPR motif structurally mimics the C-end Rule (CendR) consensus (R/K-X-X-R), a short linear motif that engages the b1 domain of neuropilin-1 (NRP1, gene NRP1) — a well-characterized cell-surface co-receptor involved in VEGF signaling, axon guidance, and notably CNS neurotrophic pathways. The Researcher correctly identified that NRP1's b1 CendR pocket, anchored by Asp320, Glu319, and Tyr297, has solved crystal structures (PDB: 2ORZ, 4DEQ) and is the binding site for peptides like EG00229 and the VEGF-C C-terminus, placing it squarely within the competence of structure-prediction pipelines like Boltz-2 and Chai-1.

The modification under test was a single L-homoarginine substitution at position 4 — the terminal arginine of the TKPR pharmacophore. Homoarginine extends the guanidinium-bearing side chain by one methylene unit (~1.5 Å additional reach) while fully preserving the +1 charge, bidentate H-bond donor geometry, and overall pharmacophoric character of the native arginine. The hypothesis was that this extended reach would allow the guanidinium to form a deeper, more stable salt bridge to Asp320 at the base of the b1 pocket, potentially yielding measurable affinity gains over native Selank. This is a well-precedented strategy: homoarginine substitutions have been used in arginine-rich peptides targeting integrin RGD pockets and other acidic recognition grooves.

Despite the scientific coherence of the hypothesis, the fold was rejected at the orchestrator's predictability gate with the reason: 'target_not_predictable: no UniProt ID resolved — target identity unconfirmed.' The Researcher anticipated this issue explicitly, noting that the canonical target list returned null IDs and requesting that the Clinical agent manually resolve NRP1 (UniProt O14786, ChEMBL CHEMBL4297). The gate, however, operates on automated ID resolution and did not accept the manual annotation in time, halting the fold before Boltz-2 or Chai-1 were invoked. No structural coordinates, pLDDT scores, or binding probability estimates were generated.

This is an important distinction: the DISCARDED verdict here reflects a pipeline infrastructure failure — specifically, the automated target-resolution step's inability to map 'Neuropilin-1 (b1 domain, C-end Rule / CendR pocket)' to a UniProt accession — rather than any structural or biological evidence against the hypothesis. NRP1 is emphatically not an obscure or unpredictable target; it has dozens of high-resolution co-crystal structures with CendR peptides, a well-defined binding pocket, and extensive medicinal chemistry literature. The scientific premise of this fold is arguably stronger than several folds that did proceed.

In the context of Selank's recent fold history at Alembic Labs, this fold represents a meaningful strategic pivot. Folds #8, #18, and #41 all pursued stability-focused modifications: C-terminal amidation (#8, PROMISING, pLDDT 0.90), N-methylation at Gly-6 (#18, PROMISING, pLDDT 0.87), and D-Thr-1 substitution (#41, PROMISING, pLDDT 0.94). Fold №72 was the first Selank fold targeting affinity enhancement at a defined molecular receptor, shifting the scientific question from 'how do we make Selank last longer in plasma' to 'can we make it bind its putative CNS target more tightly.' This is a scientifically mature progression — stability hypotheses are typically a prerequisite to affinity optimization, making this the logical next chapter.

The hArg substitution itself carries no obvious red flags. The modification is synthetically accessible (Fmoc-homoarginine(Pbf)-OH is commercially available), is not expected to alter the backbone conformation dramatically given the flexible single-bond extension, and homoarginine itself is an endogenous amino acid (found in human plasma) with a favorable safety profile. Heuristic estimates suggest negligible impact on aggregation propensity or BBB penetration relative to native Selank, and the +1 charge is preserved. The principal open question is whether the ~1.5 Å side-chain extension is a net gain (deeper pocket burial) or a net loss (over-extension past the optimal salt-bridge geometry to Asp320) — a question that requires either computational docking or direct experiment.

To rescue this fold, the simplest path is to resubmit with NRP1's UniProt accession (O14786) pre-populated in the target metadata, bypassing the automated resolution step. Alternatively, flexible-receptor docking using Glide or AutoDock Vina against PDB:2ORZ (NRP1 b1 domain with TKPR-like peptide) would provide rapid, interpretable geometry — and would complement the Boltz-2/Chai-1 pipeline rather than replace it. SPR or ITC with recombinant NRP1 b1 domain remains the gold-standard wet-lab adjudication, as used in the original EG00229 characterization studies.