Retention time mass correspondence resolution

retention-time-mass-correspondence-resolution

Summary

Match untargeted metabolomic features across two LC-MS datasets by resolving one-to-one correspondences based on m/z and retention time similarity. This skill enables cross-dataset feature alignment by computing pairwise similarity scores and selecting highest-confidence matches within specified mass and temporal tolerances.

When to use

You have two LC-MS untargeted metabolomic feature tables (each containing m/z, retention time, and intensity columns) and need to establish which features in dataset A correspond to which features in dataset B, typically for comparative metabolomics, batch effect correction, or multi-cohort feature integration.

When NOT to use

• Input datasets are targeted metabolomics (with known metabolite identities) rather than untargeted feature discovery.
• Features are already aligned (e.g., from a single LC-MS run or pre-processed common feature space).
• One or both input tables lack retention time information or have poor m/z calibration.

Inputs

• LC-MS feature table 1 (with m/z, retention time, intensity columns)
• LC-MS feature table 2 (with m/z, retention time, intensity columns)

Outputs

• Matched feature pair table (original identifiers, m/z, retention times, confidence scores)
• Feature correspondence matrix

How to apply

Load both LC-MS feature tables into Matlab, then apply the M2S matching algorithm to compute pairwise similarity scores across m/z and retention time dimensions. Define mass-to-charge and temporal tolerances to filter candidate matches and eliminate false positives. Resolve the bipartite matching problem by selecting the highest-confidence match for each feature pair, ensuring a one-to-one correspondence. Output a structured table mapping original feature identifiers, m/z values, retention times, and match confidence scores for downstream validation and reuse.

Related tools

• M2S (Matlab package implementing pairwise feature matching and one-to-one correspondence resolution via m/z and retention time similarity scoring) — https://github.com/rjdossan/M2S

Evaluation signals

• All matched pairs have m/z and retention time differences within user-specified tolerances (e.g., ±5 ppm for mass, ±0.5 min for retention time).
• Each feature in the smaller dataset is matched to at most one feature in the larger dataset (one-to-one correspondence enforced).
• Match confidence scores are monotonically ranked; the highest-scoring match is selected for each feature pair.
• Output table schema is complete: every matched pair includes original identifiers, m/z values, retention times, and confidence metrics.
• No matched pairs violate the pairwise similarity metric (e.g., Euclidean distance in normalized m/z–RT space).

Limitations

• M2S is designed for untargeted LC-MS feature matching; performance may degrade if features are poorly resolved or if m/z calibration differs significantly between datasets.
• One-to-one matching assumes no feature duplication or isobaric co-elution; overlapping or ambiguous matches may lead to arbitrary resolution.
• Sensitivity to mass and temporal tolerance thresholds; overly strict tolerances risk missing true matches, while loose tolerances risk false positives.
• No changelog or version control information was found, limiting reproducibility and compatibility tracking.

Evidence

• [readme] Matlab package to match untargeted metabolomic features of two LC-MS datasets: "Matlab package to match untargeted metabolomic features of two LC-MS datasets"
• [other] Apply pairwise feature matching by computing similarity scores across m/z and retention time dimensions using the M2S matching algorithm.: "Apply pairwise feature matching by computing similarity scores across m/z and retention time dimensions using the M2S matching algorithm"
• [other] Filter candidate matches based on mass-to-charge and temporal tolerances to eliminate false positives.: "Filter candidate matches based on mass-to-charge and temporal tolerances to eliminate false positives"
• [other] Resolve one-to-one feature correspondences by selecting the highest-confidence match for each feature pair.: "Resolve one-to-one feature correspondences by selecting the highest-confidence match for each feature pair"
• [other] Compile matched pairs into a structured output table with original feature identifiers, m/z values, retention times, and match confidence scores.: "Compile matched pairs into a structured output table with original feature identifiers, m/z values, retention times, and match confidence scores"