Work in progress (pre-release) 🚧
arteMIS is a statistically gorunded framework for robust, reproducible, and interpretable molecular networking. It is desgined to help users to tune parameters, score network quality, and compare runs by measuring how well network topology agrees with chemistry based metrics.
- Parameter tuning & benchmarking: Compare MN runs (e.g., spectral similarity metrics, diverse thresholds, min peaks, max links) with validated evaluation metrics.
- Statistcal optimization: Define optimal configuration for your MN experiments (e.g., targted mining of terpenoids).
- Topology ↔ Chemistry agreement: Quantifies how network structure (edges/nodes/components) aligns with chemistry-derived groupings or distances (e.g., morgan fingerprints and tanimoto similarity).
arteMIS does run spectral networking itself. You can evaluate the optimized configurations in your own dataset using matchMS-based metrics.
Nose hehe
Implemented (early stage):
- Component/family purity vs. chemical classes
- Neighborhood consistency vs. chemistry similarity
- Ranked run score (composite of selected metrics)
- Fingerprint-aware entanglement of topology vs. chemistry dendrograms
Planned: