Peak similarity assessment

peak-similarity-assessment

Summary

Assess spectral similarity and co-occurrence patterns among picked peaks in INADEQUATE NMR spectra to group peaks originating from the same compound. This skill operates as a clustering step after peak picking and before database matching, using spectral similarity metrics to identify peak networks.

When to use

Apply this skill when you have a set of picked peaks from INADEQUATE NMR spectra and need to group them into networks to identify which peaks co-originate from the same metabolite compound. Use it as a downstream filtering step after peak picking and before attempting to match peak networks to a metabolite database.

When NOT to use

• Peak picking has not yet been performed; apply peak picking first
• Input peaks are already grouped or pre-clustered by an external method incompatible with PyINETA's clustering module
• Spectral data does not contain co-occurrence patterns (e.g., single isolated spectra with no replicate or multi-dimensional data)

Inputs

• Picked peaks from PyINETA peak-picking module output
• Spectral data (INADEQUATE NMR spectra in PyINETA-compatible format)
• Peak coordinates and intensity values

Outputs

• Peak cluster assignments (peak-to-cluster mapping file)
• Clustered peak networks with cluster identifiers
• Network metadata documenting co-occurrence and similarity patterns
• Clustered peak data compatible with downstream matching module

How to apply

Load picked peaks from the peak-picking module output using PyINETA's picking interface. Apply the PyINETA clustering algorithm to group peaks based on spectral similarity and co-occurrence patterns—peaks that appear together across spectra and exhibit similar spectral features are clustered together. Assign each peak a unique cluster identifier and generate a mapping file documenting the peak-to-cluster assignments. Export the clustered peak data in a format compatible with downstream matching and finding modules. The clustering acts as a dimensionality reduction step that simplifies the search space for metabolite database matching by reducing noise and grouping related signals.

Related tools

• PyINETA (Implements the clustering module (pyineta.clustering) that groups peaks by spectral similarity and co-occurrence patterns) — https://github.com/edisonomics/PyINETA
• Python (Runtime environment and language for PyINETA execution)

Examples

python <path_to_pyineta_repo>/run_pyineta.py -c config.ini -s cluster -o output_dir

Evaluation signals

• Peak-to-cluster mapping file is generated and contains a valid cluster identifier for every input peak
• Cluster size distribution is reasonable and peaks within each cluster exhibit high spectral similarity scores (exact threshold depends on algorithm configuration)
• Clusters preserve co-occurrence relationships observed in raw spectra (e.g., peaks that co-appear in spectra remain in the same cluster)
• Clustered output can be successfully loaded and processed by the downstream matching module without format errors
• Manual inspection of representative clusters confirms that grouped peaks correspond to expected multiplet patterns or correlated signals from the same compound structure

Limitations

• Clustering performance depends on quality of upstream peak picking; poor peak picking produces unreliable clusters
• Algorithm assumes co-occurrence patterns are present in the spectral data; spectra with limited replication or single-scan acquisitions may produce uninformative clusters
• No changelog is publicly available for PyINETA, making it difficult to track changes in clustering algorithm behavior across versions
• Clustering operates on spectral similarity metrics without chemical structure validation; chemically unrelated peaks may cluster if they share spectral patterns

Evidence

• [intro] pyINETA implements a clustering module that filters picked peaks to identify networks of peaks from the same compound: "It is designed to filter these picked peaks to identify networks of peaks (ideally) coming from the same compound"
• [methods] Clustering is a core workflow step in the PyINETA pipeline: "Clustering

---

.. automodule:: pyineta.clustering"

• [other] Clustering operates on spectral similarity and co-occurrence patterns: "Apply the PyINETA clustering algorithm to group peaks into networks based on spectral similarity and co-occurrence patterns."
• [other] Clustering precedes matching to the metabolite database: "operating as a downstream step after peak picking and before matching to metabolite databases"
• [readme] PyINETA reads and processes INADEQUATE spectra with peak picking and clustering: "pyINETA can perform basic tasks such as reading and referencing (using basic shifting) the INADEQUATE spectra and peak picking"