Technology Partnership Proposal · 2026

AI-Driven Skin Active Discovery

Proprietary databases · Validated AI engine · Skin microbiome expertise

Guangdong-Hong Kong-Macao Greater Bay Area HKUST (GZ) Confidential
ai.mi · Confidential

ABOUT US

Who We Are

JK

Jean Krutmann

Co-Founder · Chief Scientific Advisor

World-renowned environmental dermatologist IUF–Leibniz Institut, Düsseldorf Pioneer in AhR / skin photoaging research
XJ

Jingjing Xia

Co-Founder · Chief Science Officer

MD & PhD Cell Biology, Pharmacology, Bioinformatics Skin microbiome, skin biology
LH

Leo Han

Co-Founder · Chief Technology Officer

AI & database architecture specialist Multi-agent pipeline engineering Natural product knowledge graph Full-stack AI pipeline architect
◆ Mission: AI-assisted discovery of novel skin active ingredients — small molecules, single entities, and complex formulations

TECHNOLOGY PLATFORM

End-to-End Discovery Pipeline

Data Assets AI Engine Deliverable Proprietary Databases Yunnan NP Library 200K natural products · 49K herbs 5,800 Yunnan-specific species 20K+ physical isolates TCM Knowledge Graph 7,443 formulas · 65,108 compounds 155,351 curated CP-target pairs Formula→Herb→Compound→Target→Disease EcoDermDB 3,700+ skin microorganisms 2,500+ ingredient records 70+ phenotype dimensions per organism AI-Powered Screening Pipeline GNN Screening Graph neural network · AUC 0.918 Molecular Docking + MD Simulation AutoDock Vina · Binding free energy ADMET + Toxicity Profiling Lipinski · hERG · SkinSen · HepG2 toxicity Multi-Target Scoring 200K compounds × 8,700 targets Ranked Output Ranked Candidate List Top N per target with scores Docking Scores + ADMET Binding energy · All PK parameters Safety Profile SkinSen · hERG · cytotoxicity Traditional Traceability Formula → Herb → Compound chain
【中文】我们搭建了一条端到端的 AI 活性成分发现管线。左手是数据资产：云南天然产物库、TCM 知识图谱、EcoDermDB——三个库都是自研的，竞争对手拿不到。中间是 AI 引擎：GNN 筛选、分子对接、ADMET 毒性预测、多靶点评分，全部串联成自动化 pipeline。右手是交付物：排序好的候选分子列表，每个附带对接得分、ADMET 参数、安全评估和传统溯源——可直接进入实验验证。
We've built an end-to-end AI-powered active ingredient discovery pipeline. On the left, our proprietary data assets — Yunnan NP Library, TCM Knowledge Graph, and EcoDermDB — none of which competitors can replicate. In the center, the AI engine: GNN screening, molecular docking, ADMET and toxicity profiling, and multi-target scoring, all connected in an automated pipeline. On the right, the deliverable: a ranked candidate list where every compound comes with docking scores, ADMET parameters, safety assessment, and traditional-use traceability — ready for experimental validation.

DATA ASSET 01 · YUNNAN NATURAL PRODUCT LIBRARY

Yunnan Natural Product Library

Yunnan-distinctive · Fully structured

🌻
Unmatched Chemical Diversity

Yunnan sits at the junction of 3 major biogeographic zones — covering structural space synthetic libraries cannot reach

7,000+
Formulas
traditional prescriptions
49,000
Medicinal Materials
biological resources
200,000
Natural Products
unique compounds
8,700
Targets Covered
multi-disease scope
1.1M
Activity Records
compound–target pairs
Source Composition
  • Plant-derived ~95%
  • Animal-derived (leech, snake venom)
  • Insect medicine
  • Mineral medicine
Yunnan Local Featured Data

Differentiated & Proprietary Assets

5,800+
Yunnan Medicinal Materials
12% of total · covers 云南十大云药 · rare & endemic species
3,000+
Multi-Ethnic Formulas
43% of national library · Yi, Dai, Naxi & more · 单验方 included
Physical Library
20,000+
Purified Isolates
screening-ready
+ KIB-CAS on-demand extraction
Each candidate carries centuries of empirical safety precedent (ethnopharmacology) + modern AI-predicted target affinity

DATA ASSET 02

TCM Knowledge Graph

A 5-layer proprietary network — Formula → Herb → Compound → Target → Disease · 155,351 curated compound-target associations

7,443
Formulas
→
65,108
Compounds
→
155,351
CP-Target Pairs
Proprietary Private Database

An internally maintained, continuously curated knowledge graph — not a public resource. Clients receive analysis outputs; the underlying graph remains protected.

155,351 Curated Pairs

Every compound-target association manually curated from literature — not auto-extracted. This quality bar means predictions are grounded in real experimental data, not computational noise.

Multi-target Synergy

Graph architecture enables simultaneous scoring of multi-component, multi-target interactions — capturing TCM's cocktail effect that single-target methods miss.

DATA ASSET 03 · ECODERMDB — SKIN MICROBIOME INTELLIGENCE

EcoDermDB

The only multi-layer skin microbiome knowledge base linking phenotypes, diseases, drugs & cosmetic ingredients

3,700+
Skin Microorganisms
70%+ coverage
70+
Phenotype Dims
morphology, metabolism…
700+
Skin Diseases
ICD-11 standardized
850+
Therapeutic Drugs
920 targets · 1,047 refs
1,450
Cosmetic Ingredients
21 efficacy types
2,500+
Ingredient Records
510 refs · annotated
Multi-Layer Knowledge Architecture
🦠Skin
Microorganism
→
📊Phenotype
(70+ dims)
→
🏥Skin
Disease
→
💊Drug &
Target
→
🧴Cosmetic
Ingredient
4 association types · 6-level evidence · ICD-11 · PubMed-curated
First-in-Class Features
  • 🔗4-type classification — Pathogenic / Opportunistic / Dysbiosis / Protective
  • ⭐6-level evidence grading by study design quality
  • 🔍Cross-layer query: microbe → phenotype → drug → ingredient
📊

Microbiome Impact Prediction

Predict ingredient impact on key commensals — before any clinical spend

✅

Microbiome-Friendly Claims

Evidence-graded data substantiating 'microbiome-friendly' label claims

🌿

Postbiotic Active Discovery

Cross-reference 2,500+ metabolites with skin barrier targets → shortlist

💊

Disease–Drug–Microbe Triangle

700+ diseases × 850+ drugs × 3,700+ microbes — mechanism research

AI ENGINE & COMPETITIVE MOAT

Technology Differentiation

🗄️01

Data Nobody Else Has

Proprietary TCM graph + Yunnan NP library + skin phenotype DB — not replicable from public sources

🛡️02

Ethnopharmacology Safety Buffer

Every candidate carries centuries of documented human use — dramatically reducing safety risk vs. de novo synthetic compounds

✅03

Dual Validation Logic

Every candidate carries both traditional usage precedent and modern target mechanism annotation

🦠04

Skin Microbiome Dimension

No competitor integrates skin microbiome phenotype data into active discovery

⚡05

Multi-target Synergy Engine

200K compounds × 8,700 targets full scoring matrix

APPLICATION SCENARIOS

What We Can Do for You

💡Can we discover a novel brightening molecule from nature?
→ 20–50 prioritized novel molecules with full annotation
🧬How to identify a multi-target anti-aging active?
→ Candidates ranked by synergy score
🦠Will this new ingredient disrupt the microbiome?
→ Microbiome impact report before clinical investment
🛡️Which postbiotic candidates promote barrier repair?
→ Dual-validated shortlist
✅Can natural products modulate AhR for anti-pollution?
→ ✓ Validated. Potent antagonist outperforming CH-223191
🌿What TCM herbs target hair loss pathways?
→ Multi-target TCM screening pipeline

PROOF OF CONCEPT — CASE STUDY 01

AhR-Targeted Active Discovery

Aryl Hydrocarbon Receptor (AhR) — emerging therapeutic target for atopic dermatitis and anti-pollution skin care

🤖
1
GNN Model
AUC-ROC: 0.918
→
🔍
2
Virtual Screening
30,842 NP · 736 top-scored (>0.8)
→
🎯
3
Molecular Docking
2,155 < −10 kcal/mol · best −13.04
→
💊
4
ADMET + Toxicity
All 18: Lipinski-compliant
→
🔬
5
Cell Validation
A1, A14: agonists. A9: Broad-spectrum AhR Antagonist

KEY FINDING

A Novel, Potent AhR Antagonist from Natural Products

🧪Compound A9
  • Broad-spectrum AhR antagonist
  • Potency exceeds CH-223191 at equal concentration
  • CH-223191 is the gold-standard reference compound
  • Source: Yunnan natural product library virtual screening
Why This Matters
🛡️
Anti-pollution
Blocks pollutant-induced oxidative stress
☀️
Anti-photoaging
Inhibits UV-driven AhR activation
🚀
Underexplored space
Few competitors in AhR antagonist space
✓ Prospective validation: compounds purchased blind — pipeline predicted, purchased, and confirmed.

PROOF OF CONCEPT — CASE STUDY 02

Hair Loss · Multi-Target TCM Screening

Demonstrating pipeline flexibility: from AhR to a completely different indication in one week

💧
FASN

Fatty Acid Synthase

Sebaceous gland lipid metabolism

Sebum / Oil Control
🌱
EGLN1

Prolyl Hydroxylase 2

HIF-1α → Wnt/β-catenin → hair growth

Hair Growth Cycle
⚡
LDHA

Lactate Dehydrogenase A

Dermal papilla cell energy metabolism

Follicle Energy
🔬
UAP1

UDP-GlcNAc Pyrophosphorylase

O-GlcNAc → androgen receptor activation

Androgen Signaling
⚗️
Screening Scale
65,108 compounds × 4 targets
8-step funnel: docking → ADMET → skin sensitization
Composite score: 0.4×docking + 0.3×ADMET + 0.2×drug-likeness + 0.1×(1-sensitization)
Top 30 candidates per target delivered
🌿
Headline Discovery

SC dominate FASN Top 20 — all 20 highest-ranked candidates hit FASN, with 5 of top 6 from this single genus.

Traditional use (消积瘀) aligns with lipid metabolism modulation — TCM precedent and modern mechanism are self-consistent.

PROPOSED COLLABORATION

Next Steps?

01💬
Open Discussion

Select target indication

02🎯
Priority Efficacy

Paid PoC — 4-6 weeks

03📋
Deliver List

Ranked candidates + NDA

04🔬
Bioactivity Validation

Internal or CRO

05💎
Pricing & IP

Service / co-dev / equity

🚀
4–6 Weeks

Low commitment · Max signal

Technical Service

Near-term · Quick engagement

Co-development

Mid-term · Shared IP

Strategic Investment

Long-term · Partnership