Datasets:

jerry982
/

BioFactory-Asian-Oral-Microbiome-Preview

The world's first Chinese-anchored oral microbiome multi-omics synthetic dataset — 3,872 production records, each linking genomic sequence + taxonomic abundance + clinical disease triplet (periodontal / cardiovascular / diabetes).

Built on 8× NVIDIA A800 DDP with Evo foundation model inference, anchored to 121 global WGS backbones and Chinese oral clinical seeds.

Why It Matters (3 Pain Points → 3 Solutions)

Pain Point	Our Solution
Western Bias — >80% global omics data is Caucasian-centric	100% Chinese population signal — unlock Asian market AI
Compliance Hell — real patient DNA takes years + $M to legalize	100% Synthetic — zero natural persons, GDPR / cross-border exempt
Compute Sink — 8× A800 + Evo engineering = months of sunk cost	Instant deploy — we absorbed compute; you get ready-to-train JSON-L

Audit Bulletin (OVERALL PASSED)

Dimension	Method	Result	Gate
Community Structure	PERMANOVA	p = 0.994	✓ PASS
Manifold Distance	MMD	0.060	✓ PASS
Clinical Triplet	PERMANOVA	p = 0.415	✓ PASS
Alpha Diversity	Shannon	p = 0.599	✓ PASS
Abundance Profile	Spearman	r = 1.000	✓ PASS

Synthetic n=3,872 vs Chinese reference n=30 · Cyclic Dirichlet calibration (concentration=500)

What's in the Preview (Free on Hugging Face)

38 stratified JSON-L records (preview_dataset.jsonl)
validate_preview.py — run PERMANOVA/MMD locally in 30 seconds
Full whitepaper — technical & commercial deep-dive

Preview: https://huggingface.co/datasets/jerry982/BioFactory-Asian-Oral-Microbiome-Preview

Full Commercial Access

Channel	Offering
Snowflake Marketplace	Quarterly drops / enterprise buyout
Direct License	3,872-record encrypted zip + Validation_Report.json + MD5

Contact: jerry820402@hotmail.com

中文摘要

全球首个华人锚定口腔多组学全合成数据集：3872 条三联记录（基因序列 + 菌群丰度 + 临床标签），经 Cyclic Dirichlet 流形校准，PERMANOVA p=0.994、MMD=0.060、临床三元组 p=0.415，终局审计 OVERALL PASSED。100% 合成数据，零自然人映射，天然规避 GDPR 与跨境临床数据合规壁垒。预览版 38 条已开源，完整商业授权请联系 jerry820402@hotmail.com。