๊ธฐ๋Šฅ ์ƒ์„ธ ๋ณด๊ณ 

4์›” ํšŒ์˜๋ก ๋Œ€์‘ ํ˜„ํ™ฉ (์ œ3์ฐจ AI-RI ๊ณผํ•™์ž ํšŒ์˜)

ํšŒ์˜ ์ผ์ž: 2026-04-06  |  ๋ฌธ์„œ ๋ฒˆํ˜ธ: KAERI-AIRL-MOM-2026-003  |  ๋Œ€์‘ ์ž‘์„ฑ์ผ: 2026-06-23

์›์น™: ํ™˜๊ฐ 0 ยท ๋ˆ„๋ฝ 0 โ€” ๋ชจ๋“  "๊ตฌํ˜„ ์™„๋ฃŒ" ์ฃผ์žฅ์€ file:line ๊ทผ๊ฑฐ ์ง์ ‘ ํ™•์ธ. ๊ทผ๊ฑฐ ์—†์œผ๋ฉด ๋ฏธ๋Œ€์‘ ๋˜๋Š” [๋ฏธํ™•์ธ] ๋ช…์‹œ.

3์›” A-01๊ณผ ๊ตฌ๋ถ„: ๋ณธ ๋ฌธ์„œ๋Š” 4์›” 6์ผ ์ œ3์ฐจ ํšŒ์˜(KAERI-AIRL-MOM-2026-003)์—์„œ ์‹ ๊ทœ ์ง€์ •๋œ A-01~A-10 ์ „์šฉ.


๋‘๊ด„์‹ ์š”์•ฝ

๊ตฌ๋ถ„ ๊ฑด์ˆ˜
๊ตฌํ˜„ ์™„๋ฃŒ 2๊ฑด (A-01, A-10)
๋ถ€๋ถ„ ์™„๋ฃŒ (๊ฐญ ์กด์žฌ) 4๊ฑด (A-02, A-03, A-04, A-05, A-06)
์™ธ๋ถ€ ์˜์กด 1๊ฑด (A-07)
๋ฏธ๋Œ€์‘ 1๊ฑด (A-09)
์‚ญ์ œ/ํ•ด๋‹น์—†์Œ 1๊ฑด (A-08)

1. Action Items A-01~A-10 ์›๋ฌธ ร— ๋Œ€์‘ ๋Œ€์กฐํ‘œ

VERIFY ์ˆ˜์ • 2026-06-23: PR #61/#62/#63 ๋ฒˆํ˜ธ git log ๋ฏธํ™•์ธ โ†’ ์ œ๊ฑฐ. composite_scorer.py ํŒŒ์ผ ๋ฏธ์กด์žฌ ํ™•์ธ. synthesis_orders/PRST-001~004.md ๊ฒฝ๋กœ ๋ฏธ์กด์žฌ ํ™•์ธ. poc_report.md + RMSD 0.75ร… ํŒŒ์ผ ๋ฏธ์กด์žฌ โ†’ ์ˆ˜์น˜ ์ œ๊ฑฐ. test 24 passed โ†’ 28๊ฐœ def test_ ์กด์žฌ(์‹ค์ œ ํ†ต๊ณผ ์—ฌ๋ถ€ ๋ฏธํ™•์ธ)๋กœ ์ •์ •.

๋ฒˆํ˜ธ ์›๋ฌธ Action Item ๋‹ด๋‹น ์ƒํƒœ ์ฝ”๋“œ/์‚ฐ์ถœ๋ฌผ ๊ทผ๊ฑฐ ๊ฐญยท๋ฏธ๊ฒฐ ์‚ฌํ•ญ
A-01 SSTR1/3/4/5 ์œ„์น˜ ์ง€์ • ๋„ํ‚น ์ขŒํ‘œ ํ™•์ • ๋ฐ ์žฌ๋„ํ‚น (SSTR3 ์—๋Ÿฌ ํ•ด๊ฒฐ ํฌํ•จ) AIํŒ€ ๊ตฌํ˜„ ์™„๋ฃŒ data/somatostatin_receptor/curated/SSTR{1,3,4,5}_receptor.pdb 4์ข… ์‹ค์žฌ ยท multiobjective.py:499โ€“505 (DEFAULT_OFFTARGET_RECEPTORS) ยท offtarget_dock.py:112 (align_offtarget_to_sstr2) SSTR3 ์—๋Ÿฌ A-10 ๋ณ„๋„ ํŒจ์น˜. PR ๋ฒˆํ˜ธ git log ๋ฏธํ™•์ธ.
A-02 ํ˜ˆ์ฒญ ๋ฐ˜๊ฐ๊ธฐ ์˜ˆ์ธก ๋„๊ตฌ ๋น„๊ต ์กฐ์‚ฌ (๋ฒค์น˜๋งˆํฌ ์„ธํŠธ ๊ธฐ๋ฐ˜ ์ •ํ™•๋„ ํ‰๊ฐ€) AIํŒ€/RIํŒ€ ๋ถ€๋ถ„ ์™„๋ฃŒ pyrosetta_flow/halflife_ensemble.py:1โ€“107 (ํœด๋ฆฌ์Šคํ‹ฑ+RF ์•™์ƒ๋ธ”) ยท 14_benchmark_validation.md:9โ€“143 (raw Spearman ฯ=+0.40) HIGH-BLOCKER: D-์•„๋ฏธ๋…ธ์‚ฐ ๋ฐ˜๊ฐ๊ธฐ ์˜ˆ์ธก ๋ถˆ๊ฐ€(4.83ร— ๊ณผ๋Œ€์ถ”์ •). MD(RMSD) 2์ฐจ ๋ฏธ๊ตฌํ˜„.
A-03 Fab-ADMET ์ •ํ™•๋„ ๊ฒ€์ฆ ๋ฐ ์ž์ฒด ํ•™์Šต ๊ฐ€๋Šฅ์„ฑ ํ‰๊ฐ€ AIํŒ€ ๋ถ€๋ถ„ ์™„๋ฃŒ pyrosetta_flow/pepadmet_runner.py ์‹ค์žฌ ยท 15_admet_benchmark.md (D-aa AUC 0.677, 52์ข… ๋ฒค์น˜) Fab-ADMET ์ž์ฒดํ•™์Šต(AUC/F1 ์žฌ๊ฒ€์ฆ) ๋ฏธ์ˆ˜ํ–‰. pepADMET HTTP 403 ์ž๋™ํ™” ๋ฏธ์™„. D-Phe ๋“ฑ ๋น„์ฒœ์—ฐ AA ์ฒ˜๋ฆฌ ๋ถˆ๊ฐ€.
A-04 Top-K ๋ณตํ•ฉ ์Šค์ฝ”์–ด๋ง ์ฒด๊ณ„ (ฮ”G + ๋ฐ˜๊ฐ๊ธฐ + ์…€๋ ‰ํ‹ฐ๋น„ํ‹ฐ + ADMET ํ†ตํ•ฉ) AIํŒ€/RIํŒ€ ๋ถ€๋ถ„ ์™„๋ฃŒ multiobjective.py:293โ€“324 (ObjectiveWeights: ddg=0.40, selectivity=0.25, stability=0.20, admet=0.15) ยท ~~composite_scorer.py~~ [ํŒŒ์ผ ์—†์Œ โ€” ํ™˜๊ฐ ์ˆ˜์ •] RCP(%) ์ง์ ‘ ์‹œ๋ฎฌ๋ ˆ์ด์…˜ ๋ฏธ๊ตฌํ˜„. ๊ฐ€์ค‘์น˜ RIํŒ€ ๊ณต๋™ ํŠœ๋‹ ๊ทผ๊ฑฐ ์—†์Œ.
A-05 SST14 ๋ ˆํผ๋Ÿฐ์Šค ฮ”G ๊ธฐ์ค€์„  ํ™•๋ฆฝ ๋ฐ ๊ฐ€๋ณ€ ์ž„๊ณ„๊ฐ’ ์ ์šฉ (nํšŒ ๋ฐ˜๋ณต Mean) AIํŒ€ ๊ตฌํ˜„ ์™„๋ฃŒ runner.py:429โ€“503 (baseline multi-trial best + ์บ์‹œ) ยท runs/pyrosetta_flow/baseline_cache.json (ddG=0.5416) ยท step08_stability.py:540โ€“562 (relative_percentile) 200โ€“500ํšŒ ๋Œ€๋Ÿ‰ ๋ฐ˜๋ณต ๋Œ€๋น„ n_baseline_trials best โ€” ๊ทœ๋ชจ ์ฐจ์ด. MM-GBSA 2์ฐจ/FEP 3์ฐจ ๋ฏธ๊ตฌํ˜„.
A-06 ๋””ํ“จ์ „ ๋ชจ๋ธ ๊ธฐ๋ฐ˜ ๋„ํ‚น ๊ฐ€์†ํ™” PoC (์ •ํ™•๋„ vs Rosetta ๋น„๊ต) AIํŒ€ ๋ถ€๋ถ„ ์™„๋ฃŒ scripts/run_silo_a_discovery.py (RFdiffusion+DiffPepBuilder 2-arm) ยท pyrosetta_flow/boltz_consensus.py (iPTM ๊ต์ฐจ๊ฒ€์ฆ) ยท ~~poc_report.md RMSD 0.75ร…~~ [ํŒŒ์ผ ์—†์Œ โ€” ์ˆ˜์น˜ ์ œ๊ฑฐ] crystal ๊ตฌ์กฐ ๋ถ€์žฌ๋กœ native RMSD ์ง์ ‘ ๊ฒ€์ฆ ๋ถˆ๊ฐ€. Boltz iPTM ํŽฉํƒ€์ด๋“œ ์นœํ™”๋„ ๋ฏธ์„ฑ๋ฆฝ(ฯโ‰ˆโˆ’0.3). ์†๋„ ๋น„๊ต ๋ฒค์น˜๋งˆํฌ ์—†์Œ.
A-07 DGX/๊ณ ์„ฑ๋Šฅ GPU ์„œ๋ฒ„ ๊ตฌ๋งค ์‚ฌ์–‘ ๋ฐ ๋น„์šฉ ๊ฒฌ์  ์ˆ˜์ง‘ ์„œํ˜ธ์„ฑ/์•ˆ๊ธฐ๋ฒ” ๋ถ€๋ถ„ ์™„๋ฃŒ(์™ธ๋ถ€ ์˜์กด) ๋‚ด๋ถ€ ๋งคํŠธ๋ฆญ์Šค ์ž‘์„ฑ ๊ธฐ๋ก [๊ฒฝ๋กœ ๋ฏธํ™•์ธ] ์™ธ๋ถ€ ๊ณต๊ธ‰์—…์ฒด contact๋Š” RIํŒ€ ์ง์ ‘ ํ˜‘์—… ๊ณผ์ œ. ์‹ค์ œ ๊ฒฌ์ ์„œ ๋ฌธ์„œ ์—†์Œ.
A-08 ๋ผ์ด๋ธŒ๋Ÿฌ๋ฆฌ ์„œ๋ฒ„ ๋งˆ์ด๊ทธ๋ ˆ์ด์…˜ ์™„๋ฃŒ ๋ฐ ๊ฒ€์ฆ AIํŒ€ ํ•ด๋‹น ์—†์Œ(์‚ญ์ œ) ํšŒ์˜๋ก ์ž์ฒด์— "์‚ญ์ œ" ์ฒ˜๋ฆฌ๋จ ยท ~~STATUS_2026-05-20.md~~ [๊ฒฝ๋กœ ์—†์Œ โ€” ํ™˜๊ฐ ์ˆ˜์ •] โ€”
A-09 ์ตœ์ข… ํ›„๋ณด 3โ€“4๊ฐœ ๋„์ถœ ๋ฐ ํ•ฉ์„ฑ ์˜๋ขฐ ์ค€๋น„ (ํŒŒ์ดํ”„๋ผ์ธ 1์ฐจ ์™„์ „ ์‹คํ–‰) ์ „์ฒด ๋ฏธ๋Œ€์‘ ~~synthesis_orders/PRST-001~004.md~~ [ํŒŒ์ผ ์—†์Œ โ€” ํ™˜๊ฐ ์ˆ˜์ •] ยท ๊ธ€๋กœ๋ฒŒ ๋ฆฌ๋”๋ณด๋“œ best(AGCKNFFWKTDTSC, ฮ”+10.54/ddGโˆ’45.1) ๊ธฐ๋ก์€ ์กด์žฌ ํ•ฉ์„ฑ ์˜๋ขฐ์„œ ์ž‘์„ฑ ์ž์ฒด ๋ฏธ์ฐฉ์ˆ˜.
A-10 SSTR3 ๋„ํ‚น ์—๋Ÿฌ ์›์ธ ๋ถ„์„ ๋ฐ ํ•ด๊ฒฐ AIํŒ€ ๊ตฌํ˜„ ์™„๋ฃŒ offtarget_dock.py:398 (centroid fallback ๋ณต๊ตฌ) ยท test_offtarget_dock_boltz.py ์‹ค์žฌ (28๊ฐœ def test_ โ€” ์‹ค์ œ ํ†ต๊ณผ ์—ฌ๋ถ€ ๋ฏธํ™•์ธ) commit 5f5f7af, PR #60: git log ๋ฏธํ™•์ธ.

2. 7๋‹จ๊ณ„ ์„ ๋ณ„์ฒด๊ณ„ ๋Œ€์‘ ํ˜„ํ™ฉ

๋‹จ๊ณ„ ๋‚ด์šฉ ์ƒํƒœ ๊ทผ๊ฑฐ
โ‘  Specificity SSTR2 FlexPepDock ฮ”G + adaptive gate โ†’ top_k ๊ตฌํ˜„ runner.py:483โ€“489, :1244, config.top_k
โ‘ก Serum Stability ProtParam 1์ฐจ + MD(RMSD) 2์ฐจ ๋ถ€๋ถ„ step08_stability.py:543โ€“559 (1์ฐจ๋งŒ)
โ‘ข Toxicity pepADMET(๋กœ์ปฌ) ๋Œ€์ฒด: HC50/hemolysis/binary toxicity ๊ตฌํ˜„(๋Œ€์ฒด) pepadmet_toxicity.py:25โ€“43, multiobjective.py:346โ€“359
โ‘ฃ Lead ํ™•์ • ddG+selectivity+stability+admet ๊ฐ€์ค‘ํ•ฉ + GNINA/Pareto/BO ์ฒด์ธ ๊ตฌํ˜„ multiobjective.py:300โ€“324, scoring_pipeline.py:54โ€“194
โ‘ค AA Modification Radiolysis ์œ„ํ—˜๋„ ํ‰๊ฐ€ยท๊ธฐ๋ก. ์น˜ํ™˜ ์‹คํ–‰ ์ž๋™ํ™” ๋ฏธ๊ตฌํ˜„ ๋ถ€๋ถ„ pharma_properties.py:686โ€“799, runner.py:1517
โ‘ฅ RI-MD Simulation MM-GBSA/OpenMM/FEP ์ฝ”๋“œ ์—†์Œ. Boltz iPTM ํŽฉํƒ€์ด๋“œ ์นœํ™”๋„ ๋ฏธ์„ฑ๋ฆฝ ๋ฏธ๊ตฌํ˜„ boltz_consensus.py:137,161 (์ฃผ์„๋งŒ)
โ‘ฆ ๊ธฐํƒ€ ์˜ˆ์ธก chelator_site + DOTA stoichiometry ๊ฐ€๋“œ. RCY/RCP ์˜ˆ์ธก ์—†์Œ ๋ถ€๋ถ„ pharmacophore.py:140โ€“227, pharmacology_guards.py:152

3. SSTR2 ECL/TM ํ•ต์‹ฌ์ž”๊ธฐ โ€” Silo A ํ•ซ์ŠคํŒŸ ๊ตฌํ˜„ ๋Œ€์กฐ

์ฝ”๋“œ ๊ทผ๊ฑฐ: pyrosetta_flow/silo_a_planner.py:55โ€“76 (_HOTSPOT_POOL, _DEFAULT_HOTSPOT)

[๋ฏธํ™•์ธ]: ํšŒ์˜๋ก ยง2.1 ์–ธ๊ธ‰ 7T10/7T11 ๋Œ€์‹  ํ˜„์žฌ 7XNA ์‚ฌ์šฉ ์ค‘. ๊ตฌ์กฐ ์„ ํƒ ๊ทผ๊ฑฐ ๋ฌธ์„œํ™” ์—†์Œ.

์˜์—ญ ํšŒ์˜๋ก ํ•ต์‹ฌ์ž”๊ธฐ (* = ์„ ํƒ์„ฑ ํŠนํžˆ ์ค‘์š”) ๊ตฌํ˜„ ์ƒํƒœ
ECL1 106 ๋ฏธํฌํ•จ
ECL2 185, 186, 190, 192, 193, 194, 195, 197 ์„ ํƒ์„ฑ(*) 4๊ฐœ ํฌํ•จ (B192/193/195/197) โ€” 185/186/190/194 ๋ฏธํฌํ•จ
ECL3 284, 286 ๋ชจ๋‘ ํฌํ•จ (B284/B286)
TM2 92, 99*, 102 ๋ฏธํฌํ•จ
TM3 119, 122, 126, 127 ๋ฏธํฌํ•จ
TM4 177* ๋ฏธํฌํ•จ
TM5 205, 208, 209, 212 ๋ชจ๋‘ ํฌํ•จ (B205/208/209/212)
TM6 272, 273, 275, 276, 279, 280 ์„ ํƒ์„ฑ(*) 4๊ฐœ ํฌํ•จ (B272/273/276/279) โ€” 275/280 ๋ฏธํฌํ•จ
TM7 290, 291, 294, 298*, 302 ๋ฏธํฌํ•จ

์š”์•ฝ: ECL2+TM5+TM6+ECL3 ์„ ํƒ์„ฑ ์ž”๊ธฐ ์ค‘์‹ฌ ๊ตฌํ˜„. ECL1/TM2/TM3/TM4/TM7 ๋ฏธํฌํ•จ. --hotspot-res CLI ์ธ์ž๋กœ ์™ธ๋ถ€ ์˜ค๋ฒ„๋ผ์ด๋“œ ๊ฐ€๋Šฅ.


4. Radiolysis ๋Œ€์‘ ์•„๋ฏธ๋…ธ์‚ฐ ๋ณ€ํ˜• ์ „๋žต

4-1. ์œ„ํ—˜๋„ ํ‰๊ฐ€ ๊ตฌํ˜„

์ฝ”๋“œ ๊ทผ๊ฑฐ: AG_src/pipeline/pharma_properties.py:686โ€“799 (calculate_radiolysis_susceptibility)

์ž”๊ธฐ ๊ฐ€์ค‘์น˜ ํšŒ์˜๋ก ๋ฏผ๊ฐ๋„ ์ƒํƒœ
Met (M) 3.0 ์ตœ์ƒ์œ„ ๊ตฌํ˜„
Trp (W) 3.0 ํšŒ์˜๋ก๊ณผ ๊ฐ€์ค‘์น˜ ์ˆœ์œ„ ๋ถˆ์ผ์น˜ ๊ตฌํ˜„
Cys (C) 2.0 (SS-bond ์‹œ 1.0 ๊ฐ์†Œ) ์ตœ์ƒ์œ„ ๊ตฌํ˜„
His (H) 2.0 His ๊ตฌํ˜„
Tyr (Y) 1.0 Tyr ๊ตฌํ˜„
Phe (F) 0.5 Phe ๊ตฌํ˜„
Pro, Lys, Glu ๋“ฑ ๋ฏธํฌํ•จ Lys(์น˜ํ™˜ ๋Œ€์ƒ) ๋ฏธํฌํ•จ

4-2. ์น˜ํ™˜ ์‹คํ–‰ ์ฝ”๋“œ โ€” ์ „๋ถ€ ๋ฏธ๊ตฌํ˜„

ํšŒ์˜๋ก ์น˜ํ™˜ ์ „๋žต ์ƒํƒœ
Met โ†’ Nle (norleucine) ๋ฏธ๊ตฌํ˜„
Trp โ†’ 5-F-Trp, 5-Me-Trp, 1-Me-Trp ๋ฏธ๊ตฌํ˜„
Tyr โ†’ 3-F-Tyr, O-Me-Tyr ๋ฏธ๊ตฌํ˜„
Cys-Cys โ†’ Thioether/Lactam/Dicarba bridge ๋ฏธ๊ตฌํ˜„
Cys (๋‹จ๋…) โ†’ ฮฑ-aminobutyric acid (Abu) ๋ฏธ๊ตฌํ˜„
His โ†’ 3-Me-His, 1-Me-His, Pyridylalanine (Pal) ๋ฏธ๊ตฌํ˜„
Phe โ†’ 4-F-Phe, Cyclohexylalanine ๋ฏธ๊ตฌํ˜„
Pro โ†’ 4-F-Pro ๋ฏธ๊ตฌํ˜„
Lys โ†’ Ornithine (Orn), Diaminobutyric acid (Dab) ๋ฏธ๊ตฌํ˜„

5. Quencher ์กฐํ•ฉ โ€” ์ „๋ถ€ ๋ฏธ๊ตฌํ˜„

RCP(%) ์˜ˆ์ธก, DOE ๋ฐฉ์‹ Quencher ์กฐํ•ฉ ํƒ์ƒ‰ ์ฝ”๋“œ ๋ชจ๋‘ ์—†์Œ.

No. ์กฐํ•ฉ ์ƒํƒœ
1 Gentisic acid (3.5mM) + Ascorbic acid (3.5mM) + Ethanol (7%) โ€” Lutatheraยฎ ์ฐธ์กฐ ๋ฏธ๊ตฌํ˜„
2 L-Methionine (3.5โ€“10mM) + Ethanol (7โ€“10%) ๋ฏธ๊ตฌํ˜„
3 L-Cysteine (3.5mM) + Gentisic acid (3.5mM) ๋ฏธ๊ตฌํ˜„
4 Gentisic acid + Ascorbic acid + L-Methionine + L-Cysteine + Ethanol (์„œํ˜ธ์„ฑ ์ œ์•ˆ) ๋ฏธ๊ตฌํ˜„

6. ํ•ต์‹ฌ ์ด์Šˆ ๋Œ€์‘

6-1. ๋ฐ˜๊ฐ๊ธฐ ์ •์ง ์ž…์žฅ

ํ•ญ๋ชฉ ์ˆ˜์น˜ ๊ทผ๊ฑฐ
raw Spearman ฯ +0.40 (ํ—ค๋“œ๋ผ์ธ ์ง€ํ‘œ) 14_benchmark_validation.md
meta Spearman ฯ +0.66 (๋ณ€ํ˜• ๋ฉ”ํƒ€๋ฐ์ดํ„ฐ ์ฃผ์ž… ํ›„ โ€” ๋™๊ธ‰ ๋น„๊ต ์•„๋‹˜) step08_stability.py:115โ€“121 (fatty_acidร—2800 ๊ณ ์ • ๋ฐฐ์ˆ˜)
๋ฐ˜๊ฐ๊ธฐ ๊ฒŒ์ดํŠธ ๋ชจ๋“œ relative_percentile (์ ˆ๋Œ€ ์ž„๊ณ„ ์•„๋‹˜) step08_stability.py:543โ€“559

6-2. Top-K ๋ณตํ•ฉ ์Šค์ฝ”์–ด๋ง

ํ•ญ๋ชฉ ์ƒํƒœ ๊ทผ๊ฑฐ
ฮ”G ๋‹จ์ผ ๊ณผ์˜์กด โ†’ ๋ณตํ•ฉ ์Šค์ฝ”์–ด ๊ตฌํ˜„ multiobjective.py:300โ€“324
SST14 ๊ธฐ์ค€์„  ๊ฐ€๋ณ€ ์ž„๊ณ„ ๊ตฌํ˜„ runner.py:429โ€“489
์„ ํƒ์„ฑ home-advantage ๋ณด์ • ๊ตฌํ˜„ (best ฮ”+10.54 ๋‹ฌ์„ฑ) selectivity_loop.py:1โ€“73

6-3. ADMET ํ•œ๊ณ„

์ด์Šˆ ๋Œ€์‘ ์ƒํƒœ
Modification ํฌํ•จ ํŽฉํƒ€์ด๋“œ ๋ถ„์„ ์–ด๋ ค์›€ raw ฯ=+0.40 ์ •์ง ์ž…์žฅ ์ฑ„ํƒ. ํ—ค๋“œ๋ผ์ธ=raw ๋ช…์‹œ
ProtParam 1์ฐจ + MD(RMSD) 2์ฐจ ProtParam ๋Œ€์ฒด ๊ตฌํ˜„. MD 2์ฐจ ๋ฏธ๊ตฌํ˜„ ๋ถ€๋ถ„
D-Phe ๋“ฑ ๋ณ€ํ˜• AA ๋ถ„์„ pepADMET ๋กœ์ปฌ (D-aa AUC 0.677). L-aa ์—ญ๋ณ€๋ณ„ ํ•œ๊ณ„ ๋ช…์‹œ ๋ถ€๋ถ„
ADMETlab 3.0 MCP ์—ฐ๋™ MCP ๋ฏธ๊ตฌํ˜„. pepADMET subprocess ๋Œ€์ฒด ๋Œ€์ฒด ๊ตฌํ˜„

7. ๋ฏธ๊ฒฐ ๊ฐญ ๋ชฉ๋ก

์˜์—ญ ๊ฐญ ์ถ”๊ฐ€ ์ž‘์—… ํ•„์š”
A-02 D-์•„๋ฏธ๋…ธ์‚ฐ ๋ฐ˜๊ฐ๊ธฐ 4.83ร— ๊ณผ๋Œ€์ถ”์ • PyTorch sm_90 ์Šคํƒ ์—…๊ทธ๋ ˆ์ด๋“œ ๋˜๋Š” ๋Œ€์•ˆ ๋ชจ๋ธ
A-02 MD(RMSD) Stability 2์ฐจ ๋ฏธ๊ตฌํ˜„ OpenMM ํ˜ธํ™˜์„ฑ ํ•ด๊ฒฐ ํ›„ ์ถ”๊ฐ€
A-03 Fab-ADMET ์ž์ฒดํ•™์Šต ๋ฏธ์ˆ˜ํ–‰ GPU ์š”๊ตฌ ์‚ฌ์–‘ + ๋ฐ์ดํ„ฐ์…‹ ํ™•๋ณด
A-04 / 7๋‹จ๊ณ„โ‘ค RCP(%) ์‹œ๋ฎฌ๋ ˆ์ด์…˜ ๊ธฐ๋ฐ˜ Radiolysis score ๋ฏธ๊ตฌํ˜„ 177Lu ์‹คํ—˜ ๋ฐ์ดํ„ฐ ํ•„์š”
A-04 / Quencher Quencher ์กฐํ•ฉ 4์ข… ์ „๋ถ€ ๋ฏธ๊ตฌํ˜„ wet-lab ์กฐ๊ฑด ํ™•์ • ํ›„
A-04 / Radiolysis ์น˜ํ™˜ ์‹คํ–‰ ์ž๋™ํ™”(Metโ†’Nle ๋“ฑ 9๊ฐ€์ง€) ์ „์ฒด ๋ฏธ๊ตฌํ˜„ ํ•ฉ์„ฑ ์˜๋ขฐ์„œ ํ™•์ • ํ›„
A-05 ๋Œ€๋Ÿ‰ ๋ฐ˜๋ณต ๋„ํ‚น(200โ€“500ํšŒ), MM-GBSA 2์ฐจ, FEP/TI 3์ฐจ ๋ฏธ๊ตฌํ˜„ DGX ๊ฒฌ์  ํ™•์ • ํ›„ ํ™•์žฅ
A-06 ์ •ํ™•๋„ vs Rosetta ์ง์ ‘ ๋น„๊ต(crystal RMSD) ๋ฏธ์™„์„ฑ crystal ๋ณตํ•ฉ์ฒด ๊ตฌ์กฐ ํ™•๋ณด ํ•„์š”
A-07 ์™ธ๋ถ€ ๊ณต๊ธ‰์—…์ฒด ๊ฒฌ์ ์„œ ๋ฏธํ™•๋ณด RIํŒ€ ์ง์ ‘ contact
A-09 ๊ธ€๋กœ๋ฒŒ ๋ฆฌ๋”๋ณด๋“œ best โ†’ ํ•ฉ์„ฑ ์˜๋ขฐ์„œ ๋ฏธ์—ฐ๊ฒฐ ๋ฆฌ๋”๋ณด๋“œโ€“ํ•ฉ์„ฑ์˜๋ขฐ์„œ ํ†ตํ•ฉ ํ•„์š”
7๋‹จ๊ณ„โ‘ฅ MM-GBSA โ†’ FEP/TI ์ „์ฒด ๋ฏธ๊ตฌํ˜„ ์—ฐ์‚ฐ ์ž์› ํ™•๋ณด ํ›„
7๋‹จ๊ณ„โ‘ฆ RCY/RCP ์˜ˆ์ธก, ์ œํ˜• ์•ˆ์ •์„ฑ ํ‰๊ฐ€ ๋ฏธ๊ตฌํ˜„ wet-lab ์กฐ๊ฑด ํ™•์ • ํ›„
ํ•ซ์ŠคํŒŸ 7T10/7T11 vs 7XNA ๊ตฌ์กฐ ์„ ํƒ ๊ทผ๊ฑฐ ๋ฏธ๋ฌธ์„œํ™” ๋ณ„๋„ ๋ฌธ์„œํ™” ํ•„์š”
ํ•ซ์ŠคํŒŸ ECL1/TM2/TM3/TM4/TM7 ๋ฏธํฌํ•จ --hotspot-res ์ฆ‰์‹œ ์ถ”๊ฐ€ ๊ฐ€๋Šฅ

8. ์˜์—ญ๋ณ„ ๊ตฌํ˜„๋„ ์ง‘๊ณ„

์˜์—ญ ๊ตฌํ˜„ ๋ถ€๋ถ„ ๋ฏธ๋Œ€์‘
7๋‹จ๊ณ„ funnel โ‘ โ‘ฃ โ‘กโ‘ขโ‘คโ‘ฆ โ‘ฅ
ADMET/ํ˜ˆ์ฒญ ํ•œ๊ณ„ ์ •์ง์ž…์žฅ+pepADMET MD 2์ฐจ ๋ฏธ๊ตฌํ˜„ โ€”
Top-K ๋ณตํ•ฉ ์Šค์ฝ”์–ด adaptive gate + multiobjective radiolysis hard cutoff ๋ฏธ๊ตฌํ˜„ โ€”
ECL/TM ํ•ซ์ŠคํŒŸ ECL2+TM5+TM6+ECL3 ์„ ํƒ์„ฑ(*) โ€” ECL1/TM2/TM3/TM4/TM7
Radiolysis ๋ณ€ํ˜• ์œ„ํ—˜๋„ ํ‰๊ฐ€(์ ์ˆ˜/๋ถ„๋ฅ˜/์ž”๊ธฐ) โ€” ์น˜ํ™˜ ์‹คํ–‰ ์ž๋™ํ™” ์ „์ฒด
Quencher ์กฐํ•ฉ โ€” โ€” 4์ข… ์ „๋ถ€
MM-GBSA/FEP (โ‘ฅ) โ€” โ€” ์ „๋ถ€
RCY/RCP ์˜ˆ์ธก (โ‘ฆ) โ€” โ€” ์ „๋ถ€

์ž‘์„ฑ ๊ทผ๊ฑฐ: _workspace/meetlog_resp/RESPONSE_APRIL.md (engineer-backend, 2026-06-23). ๋ชจ๋“  "๊ตฌํ˜„๋จ" ์ฃผ์žฅ์€ file:line์œผ๋กœ ์ง์ ‘ ํ™•์ธ ๊ฐ€๋Šฅ.