Figure - AI Algorithm Selection in Microbiology
Quick chooser: data shape → first algorithm family. Details in AI Algorithms in Microbiology.
flowchart TD Q[Clinical / research question] --> D{Data shape?} D -->|Gene / k-mer / SNP table| T[[Tree Ensembles in Microbiology]] D -->|Need interpretable baseline| L[[Linear and Kernel Models in Microbiology]] D -->|Images plates / stains| C[[Convolutional Neural Networks in Microbiology]] D -->|Protein or DNA sequence| Tr[[Transformers and Attention in Microbiology]] D -->|Molecule or network| G[[Graph Neural Networks in Microbiology]] D -->|No labels explore structure| U[[Unsupervised Learning in Microbiology]] D -->|Invent new sequence / molecule| Gen[[Generative Models in Microbiology]] T --> E[[Model Evaluation in Clinical Microbiology]] L --> E C --> E Tr --> E G --> E U --> H[Hypothesis only — confirm with epi / AST] Gen --> W[Wet-lab validate candidates] E --> F[[Feature Representation for Microbial ML]] F -.->|bad features / leakage| X[Stop — fix splits and labels]
How to read it
- Start from data shape, not from hype.
- Every supervised path ends at evaluation; generative ends at the bench.
- Unsupervised paths are for exploration and QC, not R/S alone.