untargeted-metabolomics-dataset-integration
Summary
Match untargeted metabolomic features across two LC-MS datasets by computing pairwise similarity scores on m/z and retention time dimensions, then resolving one-to-one correspondences with confidence filtering. This enables comparative metabolomics workflows where features from independent runs must be aligned before statistical or functional analysis.
When to use
You have two LC-MS feature tables (each with m/z, retention time, and intensity columns) from independent untargeted metabolomic experiments or replicates and need to establish one-to-one feature correspondence across them to compare abundances, detect shared metabolites, or merge datasets for downstream analysis.
When NOT to use
- Datasets are from targeted metabolomics with known internal standards and reference compounds — use targeted feature alignment instead.
- Input feature tables are not yet generated from raw LC-MS data — first perform peak picking, deconvolution, and feature table construction.
- Datasets use radically different LC-MS protocols (e.g., positive vs. negative ionization modes, different chromatographic methods) — consider post-hoc ionization or mode conversion before matching.
Inputs
- LC-MS feature table 1 (columns: m/z, retention time, intensity)
- LC-MS feature table 2 (columns: m/z, retention time, intensity)
Outputs
- Matched feature pairs table (columns: feature_id_1, feature_id_2, m/z_1, m/z_2, retention_time_1, retention_time_2, match_confidence_score)
- Unmatched features report (features with no confident correspondence)
How to apply
Load both LC-MS feature tables into Matlab. Apply the M2S matching algorithm to compute pairwise similarity scores across m/z and retention time dimensions. Filter candidate matches by applying mass-to-charge and temporal tolerances (article/README do not specify exact thresholds) to eliminate false positives. Resolve one-to-one feature correspondences by selecting the highest-confidence match for each feature pair, avoiding many-to-one or one-to-many assignments. Compile matched pairs into a structured output table with original feature identifiers, m/z values, retention times, and match confidence scores for downstream validation and analysis.
Related tools
- Matlab (Runtime environment for executing M2S pairwise feature matching algorithm on LC-MS feature tables)
- M2S (Core package implementing similarity-score-based matching of untargeted metabolomic features across two LC-MS datasets) — https://github.com/rjdossan/M2S
Evaluation signals
- All matched feature pairs have symmetric one-to-one relationships (no feature appears in multiple pairs).
- Match confidence scores are reported for each pair and fall within a reasonable range (article does not specify exact range); pairs with low confidence can be manually reviewed or filtered.
- Matched m/z and retention time pairs exhibit small deltas relative to user-defined or default tolerances, confirming that only chemically plausible correspondences were retained.
- Output table schema includes original feature identifiers from both input tables, enabling traceability back to raw feature tables and peak properties.
- Unmatched features are documented and quantified; a high proportion of unmatched features may indicate misaligned chromatography or instrument drift requiring investigation.
Limitations
- M2S is designed for pairwise matching of exactly two datasets; matching three or more datasets requires iterative or custom extensions.
- Matching accuracy depends on chromatographic reproducibility and mass accuracy of the LC-MS instrument; datasets with poor retention time or m/z stability may yield low match confidence.
- No explicit handling of adducts, isotopes, or in-source fragments is mentioned; pre-processing or separate adduct annotation may be required.
- The article provides no information on computational complexity or scalability to large feature tables (thousands of features per dataset).
- GitHub repository shows no changelog; version history and parameter tuning guidance are unavailable.
Evidence
- [other] M2S is a Matlab package designed to match untargeted metabolomic features of two LC-MS datasets.: "M2S is a Matlab package designed to match untargeted metabolomic features of two LC-MS datasets."
- [other] Load two LC-MS feature tables (each with m/z, retention time, and intensity columns) into Matlab. Apply pairwise feature matching by computing similarity scores across m/z and retention time dimensions using the M2S matching algorithm. Filter candidate matches based on mass-to-charge and temporal tolerances to eliminate false positives. Resolve one-to-one feature correspondences by selecting the highest-confidence match for each feature pair. Compile matched pairs into a structured output table with original feature identifiers, m/z values, retention times, and match confidence scores.: "Apply pairwise feature matching by computing similarity scores across m/z and retention time dimensions using the M2S matching algorithm. Filter candidate matches based on mass-to-charge and temporal"
- [intro] Matlab package to match untargeted metabolomic features of two LC-MS datasets: "Matlab package to match untargeted metabolomic features of two LC-MS datasets"
1---2name: untargeted-metabolomics-dataset-integration3description: Use when you have two LC-MS feature tables (each with m/z, retention time, and intensity columns) from independent untargeted metabolomic experiments or replicates and need to establish one-to-one feature correspondence across them to compare abundances, detect shared metabolites, or merge datasets.4license: CC-BY-4.05---67# untargeted-metabolomics-dataset-integration89## Summary1011Match untargeted metabolomic features across two LC-MS datasets by computing pairwise similarity scores on m/z and retention time dimensions, then resolving one-to-one correspondences with confidence filtering. This enables comparative metabolomics workflows where features from independent runs must be aligned before statistical or functional analysis.1213## When to use1415You have two LC-MS feature tables (each with m/z, retention time, and intensity columns) from independent untargeted metabolomic experiments or replicates and need to establish one-to-one feature correspondence across them to compare abundances, detect shared metabolites, or merge datasets for downstream analysis.1617## When NOT to use1819- Datasets are from targeted metabolomics with known internal standards and reference compounds — use targeted feature alignment instead.20- Input feature tables are not yet generated from raw LC-MS data — first perform peak picking, deconvolution, and feature table construction.21- Datasets use radically different LC-MS protocols (e.g., positive vs. negative ionization modes, different chromatographic methods) — consider post-hoc ionization or mode conversion before matching.2223## Inputs2425- LC-MS feature table 1 (columns: m/z, retention time, intensity)26- LC-MS feature table 2 (columns: m/z, retention time, intensity)2728## Outputs2930- Matched feature pairs table (columns: feature_id_1, feature_id_2, m/z_1, m/z_2, retention_time_1, retention_time_2, match_confidence_score)31- Unmatched features report (features with no confident correspondence)3233## How to apply3435Load both LC-MS feature tables into Matlab. Apply the M2S matching algorithm to compute pairwise similarity scores across m/z and retention time dimensions. Filter candidate matches by applying mass-to-charge and temporal tolerances (article/README do not specify exact thresholds) to eliminate false positives. Resolve one-to-one feature correspondences by selecting the highest-confidence match for each feature pair, avoiding many-to-one or one-to-many assignments. Compile matched pairs into a structured output table with original feature identifiers, m/z values, retention times, and match confidence scores for downstream validation and analysis.3637## Related tools3839- **Matlab** (Runtime environment for executing M2S pairwise feature matching algorithm on LC-MS feature tables)40- **M2S** (Core package implementing similarity-score-based matching of untargeted metabolomic features across two LC-MS datasets) — https://github.com/rjdossan/M2S4142## Evaluation signals4344- All matched feature pairs have symmetric one-to-one relationships (no feature appears in multiple pairs).45- Match confidence scores are reported for each pair and fall within a reasonable range (article does not specify exact range); pairs with low confidence can be manually reviewed or filtered.46- Matched m/z and retention time pairs exhibit small deltas relative to user-defined or default tolerances, confirming that only chemically plausible correspondences were retained.47- Output table schema includes original feature identifiers from both input tables, enabling traceability back to raw feature tables and peak properties.48- Unmatched features are documented and quantified; a high proportion of unmatched features may indicate misaligned chromatography or instrument drift requiring investigation.4950## Limitations5152- M2S is designed for pairwise matching of exactly two datasets; matching three or more datasets requires iterative or custom extensions.53- Matching accuracy depends on chromatographic reproducibility and mass accuracy of the LC-MS instrument; datasets with poor retention time or m/z stability may yield low match confidence.54- No explicit handling of adducts, isotopes, or in-source fragments is mentioned; pre-processing or separate adduct annotation may be required.55- The article provides no information on computational complexity or scalability to large feature tables (thousands of features per dataset).56- GitHub repository shows no changelog; version history and parameter tuning guidance are unavailable.5758## Evidence5960- [other] M2S is a Matlab package designed to match untargeted metabolomic features of two LC-MS datasets.: "M2S is a Matlab package designed to match untargeted metabolomic features of two LC-MS datasets."61- [other] Load two LC-MS feature tables (each with m/z, retention time, and intensity columns) into Matlab. Apply pairwise feature matching by computing similarity scores across m/z and retention time dimensions using the M2S matching algorithm. Filter candidate matches based on mass-to-charge and temporal tolerances to eliminate false positives. Resolve one-to-one feature correspondences by selecting the highest-confidence match for each feature pair. Compile matched pairs into a structured output table with original feature identifiers, m/z values, retention times, and match confidence scores.: "Apply pairwise feature matching by computing similarity scores across m/z and retention time dimensions using the M2S matching algorithm. Filter candidate matches based on mass-to-charge and temporal"62- [intro] Matlab package to match untargeted metabolomic features of two LC-MS datasets: "Matlab package to match untargeted metabolomic features of two LC-MS datasets"
HolobiomicsLab