[cschanhniem]

Summary

• c_amidation_recommended (bool) and c_amidation_reason (str) added to SynthQC dataclass and to_dict() output.
• check_sequence() now evaluates whether C-terminal amidation would benefit the candidate and populates a C_AMIDATION_RECOMMENDED flag in the flags list.
• 11 new tests in TestCTerminalAmidation covering all recommendation cases.

Scientific rationale

~70% of natural helical AMPs that reach clinical trials are synthesized with C-terminal amidation (CONH₂ instead of COOH). The modification:

1. Charge: adds ~+0.5 to +1.0 net charge at pH 7.4 (replaces anionic carboxylate with neutral amide); removes the negative contribution from the C-terminal COOH
2. Serum stability: removes carboxypeptidase cleavage site at C-terminus → extends effective half-life
3. Membrane binding: additional positive charge enhances electrostatic attraction to negatively charged bacterial membranes
4. Cost: zero additional cost in most SPPS workflows — simply specify "C-terminal amide" or "CONH₂" in the synthesis order form

Impact on pilot panel

Primary beneficiary: SEED-004 variants (sequence ALPFIGRVLSGIL + variants):

• Current charge_pH7.4 ≈ 0.8 (just above LOW_CHARGE threshold)
• C-terminus = L (leucine, neutral)
• Post-amidation: effective charge ≈ 1.3–1.5 → above LOW_CHARGE, better membrane interaction
• Combined with serum_stability=0.85 (best in panel), this makes SEED-004 variants even stronger candidates

Other sequences ending in G, A, L, I, V, M, F, W with charge < 3.0 also flagged.

Recommendation logic

if c_term not in (K, R) and charge_pH7.4 < 3.0:
    → c_amidation_recommended = True  (strong: cost-free, meaningful improvement)
elif c_term not in (K, R):
    → c_amidation_recommended = False  (optional: mention in reason string)

Test plan

• pytest tests/test_presynth_qc.py — 74/74 pass
• Full suite — 883/883 pass (pre-push hook verified)
• SEED-004 scaffold (ALPFIGRVLSGIL) → c_amidation_recommended=True
• K-terminus sequence → c_amidation_recommended=False
• R-terminus sequence → c_amidation_recommended=False
• to_dict() exposes both c_amidation_recommended and c_amidation_reason