Metabolite feature correspondence validation

metabolite-feature-correspondence-validation

Summary

This skill identifies high-confidence anchor feature pairs across two disparately-acquired LC-MS metabolomics datasets by applying retention time windows, m/z tolerance, and abundance thresholds to establish a validated correspondence basis for subsequent RT mapping and alignment scoring. It is essential for bridging non-identical instrumental and acquisition conditions in untargeted metabolomics.

When to use

Apply this skill after m/z grouping and pairwise alignment detection when you have a metabCombiner object containing candidate feature pair alignments and need to select a subset of mutually abundant, high-confidence anchors to anchor a nonlinear retention time mapping spline. Specifically, use it when your input datasets exhibit systematic RT drift or offset and you require validated reference points (anchor pairs) with both strong m/z–RT proximity and relative abundance concordance.

When NOT to use

• Input datasets have already been aligned by a prior orthogonal method and require no further anchor validation.
• You lack abundance measurements or relative quantification data across both datasets; anchor selection requires tolQ (relative abundance tolerance) evaluation.
• Datasets were acquired under identical instrumental and LC conditions with negligible RT drift; anchor selection is unnecessary if all feature pairs are already well-aligned in RT space.

Inputs

• metabCombiner object (post-m/z grouping and pairwise alignment detection)
• Candidate feature pair alignments with m/z, retention time, and abundance annotations

Outputs

• Selected anchor feature pair set (table with rtx and rty retention time columns)
• CSV export of anchor pairs (optional)

How to apply

Load the metabCombiner object from the prior m/z grouping and pairwise alignment step. Call selectAnchors() with window parameters (windx and windy defining retention time windows around anchor points in the X and Y datasets), m/z tolerance (tolmz), relative abundance tolerance (tolQ), and linear retention time quantile difference tolerance (tolrtq). Smaller window values increase anchor count but include more outliers; balance specificity against RT mapping robustness. Set useID=FALSE to rely on similarity scores rather than compound IDs. Extract the validated anchor set using getAnchors(), which returns a table with rtx and rty retention time columns. The anchor set serves as the ground truth for the subsequent RT mapping spline that will align all remaining feature pairs.

Related tools

• metabCombiner (R package providing selectAnchors() and getAnchors() methods for anchor-based LC-MS metabolomics dataset alignment) — https://github.com/hhabra/metabCombiner
• mgcv (R package providing the generalized additive model (gam) function used by metabCombiner for fitting nonlinear RT mapping splines anchored to validated feature pairs)

Examples

# Load metabCombiner object 'mc' from prior m/z grouping step
mc <- selectAnchors(mc, windx=0.03, windy=0.02, tolQ=0.3, tolmz=0.003, tolrtq=0.3, useID=FALSE)
anchors_df <- getAnchors(mc)
write.csv(anchors_df, file='selected_anchors.csv', row.names=FALSE)

Evaluation signals

• Anchor set size and composition: number of selected anchors should be substantial enough (~20–50+) to support robust spline fitting, without excessive outliers; validate by visual inspection of rtx vs. rty scatter plot.
• m/z and RT concordance: all anchors must satisfy tolmz (m/z tolerance, typically ~0.003 Da) and rtq (linear RT quantile difference, typically ≤0.3); verify no anchors fall outside these thresholds.
• Relative abundance agreement: anchors must satisfy tolQ (relative abundance tolerance, typically ~0.3); check that paired feature abundances are reasonably similar, indicating genuine metabolite correspondence.
• Spline fit quality: after RT mapping, plot observed rtx against predicted rty; residuals should be small and randomly distributed, confirming anchors faithfully represent the underlying RT transformation.
• Downstream alignment precision: verify that the RT-mapped feature pairs exhibit improved cosine similarity scores and lower false-positive alignment rates in the subsequent Feature Pair Alignment Scoring step.

Limitations

• Smaller retention time window values (windx, windy) increase anchor count but introduce outliers that can degrade spline fit quality; requires manual parameter tuning and validation.
• Anchor selection depends on abundance data quality; datasets with high missingness or poor peak-picking reproducibility may yield spurious anchors with inflated relative abundance concordance by chance.
• Linear RT quantile difference tolerance (tolrtq) assumes approximately linear drift; datasets with strongly nonlinear RT shifts may not be well-served by anchor pairs selected under this metric, requiring alternative RT mapping strategies.
• The selectAnchors() function does not validate that selected anchors span the full RT range of both datasets; anchors clustered in a single RT region may provide poor extrapolation for features at extremes.

Evidence

• [other] The selectAnchors function identifies anchor feature pairs by applying retention time windows (windx and windy around anchor points), m/z tolerance (tolmz), relative abundance tolerance (tolQ), and linear retention time quantile difference tolerance (tolrtq): "The selectAnchors function identifies anchor feature pairs by applying retention time windows (windx and windy around anchor points), m/z tolerance (tolmz), relative abundance tolerance (tolQ), and"
• [other] smaller window values increase the number of anchors including outliers, and results are retrieved via getAnchors method returning a table with rtx and rty columns: "smaller window values increase the number of anchors including outliers, and results are retrieved via getAnchors method returning a table with rtx and rty columns"
• [other] Apply selectAnchors with window parameters (windx=0.03, windy=0.02), tolerance thresholds (tolQ=0.3, tolmz=0.003, tolrtq=0.3), and useID=FALSE to identify high-confidence anchor feature pairs based on similarity scores and m/z–RT proximity: "Apply selectAnchors with window parameters (windx=0.03, windy=0.02), tolerance thresholds (tolQ=0.3, tolmz=0.003, tolrtq=0.3), and useID=FALSE to identify high-confidence anchor feature pairs"
• [intro] metabCombiner determines possible feature pair alignments (FPAs) and determines their validity through a pairwise similarity score: "metabCombiner determines a list possible feature pair alignments (FPAs) and determines their validity through a pairwise similarity score"
• [intro] Feature alignment between datasets acquired under non-identical conditions presents numerous opportunities in untargeted metabolomics. The key challenge is achieving a correspondence between: "Feature alignment between datasets acquired under non-identical conditions presents numerous opportunities in untargeted metabolomics"