1️⃣ DMCausalTriangulate-Kor DM · T2DM 약물 × 20+ outcome 5-design causal triangulation
디엠코절트라이앵귤레이트코어 — T2DM 신약(SGLT2i·GLP-1RA·tirzepatide·DPP-4i·metformin·신흥 orforglipron·VK2735·survodutide·retatrutide) × 20+ outcome(macro-CV: CV death·MI·stroke·HHF·UA / micro: DKD·DR·DPN·DFU·CKD-progression·ESKD / non-CV: 12개 obesity-related cancer·all-cause·인지/Alzheimer's·sarcopenia·골절·UTI/Fournier·DKA·pancreatitis·MTC·suicidality) 격자에 대해 5-design(RCT·observational·target-MR·ex vivo·within-subject) effect grid + concordance score + Lawlor 2016 IJE 5-criterion 자동 채점 + discordance bias direction 자동 진단(confounding by indication·healthy-user·survivor·detection·protopathic·canalization·pleiotropy·dose-translation·external validity taxonomy) + 7가지 triangulation-targeted 후속 design 카드(active comparator new-user·target-MR pQTL IV·negative control outcome·2x2 factorial RCT·N-of-1 crossover·bariatric natural experiment·bridging ex vivo)를 한국어로 산출하는 standalone Streamlit T2DM cross-design causal robustness 도구
T2DM 약물-outcome 쌍 ≥100쌍(약물 class 5개 + 신흥 × outcome 20+개). EMPA-REG·CANVAS·DECLARE·CREDENCE·DAPA-CKD·DAPA-HF·DELIVER·EMPEROR-Reduced/Preserved·SOUL·FLOW·LEADER·SUSTAIN-6·REWIND·EXSCEL·PIONEER 6·SURPASS-CVOT·STEP-HFpEF/HFrEF·SAVOR-TIMI·EXAMINE·TECOS·CARMELINA·UKPDS·DPP/DPPOS 등 50+ RCT. Medicare D 1.5억·CPRD UK 1,500만·국민건강보험 NHIS-NSC 100만. SLC5A2(SGLT2)·GLP1R·DPP4·HMGCR(metformin proxy)·BMI MR target-MR 활성 영역. 단일 design 의존 위험: SGLT2i CV/신장 benefit은 5-design 모두 concordance(robust)·DPP-4i와 pancreatitis는 RCT+MR null·observational만 signal(confounding by indication+detection bias)·metformin과 cancer는 observational+/RCT short-term null/MR weak(healthy-user bias)·GLP-1RA와 Alzheimer's는 observational+/MR weak/RCT EVOKE ongoing(미해결). cross-design triangulation grid + discordance 진단 + triangulation-targeted 후속 가설 자동 도구 0. (1) 5-design effect ingest + ATC × MeSH 격자 정규화 (2) 5-design triangulation grid + concordance score + Lawlor 5-criterion (3) discordance bias direction taxonomy 자동 진단 (4) 7가지 후속 design 카드 + 표본수·추적기간·예상 가설 (5) HTA·규제·OpenClaw·닥터앤서 리포트 docx/PDF. OpenClaw 바이오 피벗 → 당뇨 약물 재조합 후보 target validation·off-target risk·long-term outcome 예측 layer. 닥터앤서 3.0 → 당뇨 sub-task robust endpoint priority list layer.
main --grid
main --discordant --top 5
로딩...
로딩...
2️⃣ ObesityTriangulate-Kor Obesity · 항비만 10 intervention × 25+ outcome 6-design + weight-loss mediator 분리
오베시티트라이앵귤레이트코어 — 항비만 intervention 10종(GLP-1RA semaglutide·liraglutide / GIP-GLP-1 dual tirzepatide / triple agonist retatrutide / CagriSema·orforglipron / bariatric RYGB·SG·OAGB·SADI·DJB / lifestyle intensive CR·DASH·Mediterranean·ketogenic·IF) × 25+ outcome(CV death·all-cause·12개 IARC obesity-related cancer·T2DM 발생·MASH 회귀·F2-F3 회귀·sarcopenia·골절·인지/dementia·우울·임신/산후·OSA·OA·HFpEF/HFrEF·CKD) 격자에 대해 6-design(RCT·observational·BMI/WHR/body fat % MR·bariatric natural experiment·within-subject crossover·DIO animal) effect grid + concordance + Lawlor 5-criterion + multivariable MR(BMI·WHR·body fat % 동시 IV) 로 adiposity 표현형 분리 + mediation analysis 로 weight-loss-mediated vs weight-loss-independent 효과 분리 + bariatric preference-IV 2SLS ratio estimator + bias taxonomy(selection bias for bariatric·motivation bias for lifestyle 포함) + 8가지 후속 design 카드(active comparator·MVMR·negative control·2×2 factorial·crossover·bariatric IV·DIO weight-matched·mediation N-of-1)를 한국어로 산출하는 standalone Streamlit anti-obesity cross-design causal robustness 도구
항비만 intervention-outcome 쌍 ≥250쌍. SELECT(semaglutide CV in non-DM obesity N=17,604)·SURMOUNT-MMO(tirzep CVOT N=15,000+)·STEP-HFpEF/HFrEF·ESSENCE(semaglutide MASH)·SURPASS-CVOT·MERIDIAN·TRIUMPH·REDEFINE 1/2/3·SURMOUNT-1~5/MMO/teens·STEP 1~10·BURN·LIBERA(survodutide) 50+ RCT. SOS Swedish Obese Subjects(매우 long-term)·STAMPEDE·LABS·NHIS 비만수술 코호트·Mayo·Cleveland Clinic. BMI MR(SNP 941+)·WHR(346+)·body fat %(76+)·visceral·subcutaneous adipose·multivariable MR + proteome-MR(leptin·adiponectin·GDF15) 활성. 핵심 분리 문제: 체중감량 자체가 outcome 매개인지 vs intervention의 weight-loss-independent 효과인지. SELECT semaglutide CV benefit가 weight-loss-mediated 80%인지·tirzep sarcopenia가 lean mass loss attribution인지·bariatric cancer↓가 weight-loss-mediated인지·IF healthy-user bias인지·lifestyle motivation bias인지. (1) 6-design effect ingest + ATC × IARC 12 cancer × MeSH 격자 정규화 (2) 6-design grid + concordance + MVMR + mediation(weight-loss-mediated vs direct) + Lawlor 5-criterion (3) bias taxonomy 자동 진단 + bariatric preference-IV 2SLS ratio estimator 시뮬레이션 (4) 8가지 후속 design 카드 + 표본수·primary endpoint·MVMR instrument 요구 (5) HTA·규제·KASMBS·OpenClaw docx/PDF. OpenClaw 바이오 피벗 → 비만 후보 weight-loss-mediated vs direct 분리·target validation layer. KASMBS HTA·KSSO·KOSSO·KSPED cancer prevention/sarcopenia/MASH 회귀 evidence robustness session 표준 dashboard.
main --summary
main --mediation semaglutide "CV death"
로딩...
로딩...
3️⃣ MASLDTriangulate-Kor MASLD · 단계×15+ outcome 5-design + PNPLA3 genotype-stratified + 한국 lean MASLD
매슬드트라이앵귤레이트코어 — MASLD 단계(steatosis·MASH·F2·F3·F4·biomarker FIB-4/VCTE/MRI-PDFF/ELF) × 15+ outcome(LRC mortality·HCC·decompensation·CV death·all-cause·CKD/ESKD·12개 extra-hepatic cancer·T2DM 발생·sarcopenia·인지/dementia) 격자에 대해 5-design(observational NHANES·UK Biobank·KoNEHS·NHIS-국민건강보험·MR PNPLA3 rs738409/TM6SF2 rs58542926/HSD17B13 rs72613567/MBOAT7/GCKR/proteome-MR·RCT post-resmetirom MAESTRO-NASH-OUTCOMES interim/ESSENCE/SYNERGY-NASH/ENLIGHTEN/SYMMETRY·ex vivo PCLS/DIO/HFD/MCD animal·within-subject lifestyle crossover) effect grid + concordance + Lawlor 5-criterion + multivariable MR(PNPLA3 liver-specific + BMI + T2DM 동시 IV) 로 MASLD-mediated vs metabolic-confounder-mediated 분리(LIVER_SPECIFICITY weight 기반 decomposition) + TM6SF2 MASLD↑·CV↓ discordant 패턴 자동 flag + PNPLA3 GG/GT/TT genotype-stratified treatment effect 가설 (MAESTRO-NASH·ESSENCE subgroup·한국 G allele AF 0.30 적용) + 한국 lean MASLD subphenotype 별도 layer + bias taxonomy(metabolic syndrome confounding·shared genetic confounder·reverse causation MASLD cause vs marker·ancestry external validity·detection bias for NIT 포함) + 8가지 후속 design 카드(MVMR PNPLA3+BMI+T2DM·PCLS patient-derived genotyped·active comparator new-user·genotype-stratified RCT subgroup·within-subject lifestyle crossover with PNPLA3·negative control·bridging animal→human·bariatric MASLD natural experiment)를 한국어로 산출하는 standalone Streamlit MASLD cross-design causal robustness 도구
MASLD-outcome 쌍 ≥75쌍. NHANES III 30년 추적·UK Biobank N=500,000·KoNEHS N>40,000·국민건강보험 N>1,000만. PNPLA3 rs738409 G allele freq 23-49%(한국 ~30%)·MASLD risk OR 2-3·HCC OR 12 in cirrhosis·TM6SF2 rs58542926 MASLD↑·CV↓ cleave-and-rebuild 가설·HSD17B13 rs72613567 loss-of-function ↓MASLD↓HCC protective·MBOAT7·GCKR·proteome-MR(FGF21·adiponectin·GDF15) 활성. MAESTRO-NASH(resmetirom 2024 가속승인)·VOYAGE(VK2809)·ENLIGHTEN(pegozafermin)·SYMMETRY(efruxifermin)·SYNERGY-NASH(tirzep)·ESSENCE(sema)·LIBERA(survodutide) phase 2/3 폭증. 핵심 분리 문제: MASLD가 outcome의 원인인지 vs 공통 metabolic syndrome confounder의 marker인지. observational MASLD-CV mortality 강한(+)이지만 PNPLA3 MR null·TM6SF2 MR ↓CV(MASLD↑·CV↓ discordant) → metabolic confounder 우세 가능성. MASLD-extra-hepatic cancer는 PNPLA3 MR weak. F3/F4-LRC/HCC는 PNPLA3+HSD17B13 strong → MASLD 직접 인과. (1) 5-design effect ingest + 단계 × biomarker × MeSH 격자 (2) MVMR(PNPLA3+BMI+T2DM 동시 IV)로 MASLD-mediated vs metabolic 분리 + TM6SF2 discordant 자동 flag (3) PNPLA3 GG/GT/TT subgroup treatment effect 가설 + HSD17B13 protective allele modification + 한국 ancestry AF 적용 (4) 8가지 후속 design 카드 + MR instrument 요구·genotype stratification (5) KASL·AASLD/EASL·OpenClaw·MASLDStager docx/PDF. OpenClaw 바이오 피벗 → hepatic 후보 genotype-precise target validation layer. 핀포인트 MASLDStager → staging-outcome robustness layer. KASL/AASLD/EASL 가이드라인 위원회 cross-design causal robustness layer.
main --grid
main --mediation MASH "CV death"
로딩...
로딩...