1. Our Commitment
GenoMatrix is committed to operating with the highest standards of transparency. We believe that in the field of genetic data, transparency is not only an ethical obligation but also the cornerstone of scientific credibility. This report regularly discloses our data practices, governance structure, and service status.
2. Data Scale and Sources
2.1 Database Overview
- Reference Genome: EquCab3.0 (GCF_002863925.1), coordinate-frozen
- Breed Coverage: 400+ equine breeds (including all Chinese native breeds)
- Total Data Volume: 6.8B rows, 824 data tables
- Genomic Variants: 40M+ SNPs/InDels, VEP-annotated 87.8M rows
- Causal Variants: 49,697 TIER-classified loci (TIER 1-5)
- OMIA Phenotypes: 328 equine Mendelian phenotypes
- Gene Networks: 11.4M protein interaction edges (StringDB)
- Literature Library: 150K+ indexed research papers
2.2 Data Sources
- Public Databases: NCBI dbSNP, Ensembl, OMIA, UCSC Genome Browser
- Partner Institutions: Genotyping data from domestic equine conservation farms and research institutes
- Proprietary Pipeline: JumpChain™ 7-stage, 16-step data refinement pipeline
- Literature Mining: NLP extraction and structuring from 150K+ papers
3. Data Quality and Validation
3.1 Evidence Grading System (TIER)
| Tier | Criteria | Count |
|---|
| TIER 1 | OMIA-confirmed + multi-literature validation + functional experimental evidence | Core pathogenic loci |
| TIER 2 | OMIA-confirmed + literature support | High-confidence causal variants |
| TIER 3 | GWAS significant association + functional prediction support | Candidate causal variants |
| TIER 4 | Single literature report + computational prediction | Associations pending validation |
| TIER 5 | Computational prediction or preliminary association signal | Exploratory markers |
3.2 Validation Status
- Coordinate System: Locked to EquCab3.0; modifications require Ensembl REST validation + PMC literature citation
- rsID: NCBI dbSNP as authoritative source; unverified loci marked "UNVERIFIED"
- Functional Prediction: NCDS v3.4 framework is "hypothesis-driven", not clinically validated conclusions
4. NCDS Athletic Performance Assessment Model
4.1 Model Architecture
- Core Layer: 4 core genes (DMRT3/BDNF/CHRNA1/SCN4A)
- N Layer: 6 sub-layers, 57 genes (neural-skeletal-developmental pathways)
- H Layer: 4 sub-layers, 41 genes (higher-order regulatory networks)
- Total: 102 genes, 4 roles, 6 validation chains
4.2 Limitations Statement
- NCDS scores are based on statistical associations of genetic markers, not direct measurements of athletic performance
- Environmental factors (training, nutrition, management) have significant impact on athletic performance
- The model is under continuous iteration; current version is v3.4
- Results are for breeding reference only and do not constitute commercial value commitments
5. AI Model Transparency
- EQAI Large Model: Fine-tuned on equine biology corpus; training data is publicly available
- HorseBERT: v1.0 production version frozen, using projection head retraining strategy
- Data Provenance: Model outputs are traceable to ELKG evidence sentences and original literature
- Known Biases: English literature coverage is stronger than Chinese literature; Thoroughbred data is more abundant than local breed data
6. Security Audits
- HTTPS/TLS 1.3 end-to-end encryption
- Layered database access control
- Genetic data stored in desensitized form
- Regular security assessments (most recent: Q2 2026)
7. Governance Structure
- Operating Entity: GenoMatrix Technology (Shanghai) Co., Ltd.
- Data Officer: Liu Wenzheng (Executive Director)
- Scientific Advisory: Expert committee in the field of equine genomics
- Audit Cycle: Annual report with quarterly internal reviews
8. Contact