🧫 BactAI-D — Microbiology Phenotype Identification
Database updated: 2025-12-30
BactAI-D is a schema-driven microbiology identification system that combines deterministic phenotype parsing, an extended laboratory test schema, a genus-level machine learning classifier, and retrieval-augmented generation (RAG) to provide evidence-grounded genus interpretation and a structured decision aid. (First Analysis may take 30 seconds)
Top 5 Genus Predictions (Decision Table)
Supported Phenotype Fields (Core Schema)
This page summarizes the core fields currently supported by the deterministic parsers and the unified scoring engine. Only recognized fields influence scoring; unrecognized descriptors are retained in raw text but not used for structured matching. BactAI-D is capable of extending it's own schema via testing phases. These are documented and handled by the Trifusion model of parsing.
1) Gram / Morphology
- Gram stain: Positive, Negative, Variable, Unknown
- Shape: Cocci, Bacilli, Rods, Short Rods, Yeast, Spiral, Variable, Unknown
2) Oxygen & Motility
- Oxygen requirement: Aerobic, Anaerobic, Facultative, Microaerophilic, Unknown
- Motility: Positive, Negative, Variable, Unknown
- Motility type (if provided): Peritrichous, Polar, Tumbling, Swarming, Unknown
3) Colony / Growth
- Colony morphology: free-text descriptors (e.g., “Small; Translucent; Smooth”)
- Colony pattern: Smooth, Rough, Mucoid, Dry, Variable, Unknown
- Pigment: Positive / Negative (or specific pigment text if your schema supports it)
- Odor: None / specific odor text / Unknown
- Haemolysis: Positive / Negative and type (Alpha/Beta/Gamma) if present in input
4) Core Biochemistry (examples)
- Catalase, Oxidase, Indole, Urease
- Citrate, Methyl Red, VP
- H2S
- Nitrate reduction
- Lysine decarboxylase, Ornithine decarboxylase, Arginine dihydrolase
- Esculin hydrolysis, Gelatin hydrolysis, DNase
- ONPG
- NaCl tolerance (where encoded)
5) Carbohydrate Utilisation (examples)
- Glucose fermentation, Lactose fermentation, Sucrose fermentation
- Additional sugars where present in your extended schema (e.g., xylose, rhamnose)
If you want, we can later generate this tab dynamically from your extended_schema.json so it always matches your schema exactly.
Evaluate parsers, train from gold tests, tune parser weights, train the genus-level model, build the RAG index, and commit artefacts back to the HF Space repository.
Built by Zain Asad