Blank Contamination Filtering

blank-contamination-filtering

Summary

Remove MS1 features that are overrepresented in blank control samples (solvent blanks, media blanks) using group-based relative abundance thresholds. This skill corrects for carryover contamination and background signals that confound biological signal in metabolomics peak tables.

When to use

Apply this skill when your peak table includes features flagged in blank control samples (e.g., solvent blanks, media blanks) at relative abundance above a project-specific threshold. Use it after importing raw peak tables and before applying replicability or statistical filters, especially when carryover between samples or media/solvent contamination is suspected.

When NOT to use

• Input peak table is already pre-filtered at the instrument level without blank controls recorded.
• No blank samples were collected or included in the experimental design; filtering requires a reference blank group.
• All features in blank samples are technical replicates of biological samples; applying group filter would remove authentic biology.

Inputs

• mpactr object (R6 reference object post-import_data())
• peak intensity table (Progenesis CSV format or equivalent)
• sample metadata with group/batch annotations identifying blank controls

Outputs

• filtered mpactr object with blank-contaminated features marked/removed
• data.table from qc_summary() indicating filter status per feature (passed or failed 'group' filter)

How to apply

Call filter_group() on your mpactr object, specifying the blank sample group to remove (e.g., 'Solvent_Blank' or 'Media_Blank') and setting a relative_abundance_threshold (typical range 0.01–0.05, or 1–5%). The filter compares feature intensity in blank samples to biological samples and flags/removes features where blank intensity exceeds the threshold relative to sample intensity. Rationale: blank-dominated features are likely artifacts (carryover, leached solvents, media components) rather than authentic metabolites. Repeat filter_group() for each blank type in your study design (solvent blank, then media blank). The order matters: remove solvent contamination first, then media, to avoid masking media-only contaminants. After filtering, inspect the qc_summary() output to confirm the number of features removed per blank filter and verify that authentic biological signals are retained.

Related tools

• mpactr (R6 package hosting filter_group() method for blank-group-based feature removal with in-place reference semantics) — https://github.com/mums2/mpactr
• data.table (Underlying tabular format returned by qc_summary() for inspection and downstream analysis of filter results)
• R (Language and runtime for mpactr package and filter_group() execution)

Examples

library(mpactr); obj <- import_data('cultures_peak_table.csv', 'cultures_metadata.csv', format='Progenesis'); obj$filter_group(group_to_remove='Solvent_Blank', relative_abundance_threshold=0.01); obj$filter_group(group_to_remove='Media_Blank', relative_abundance_threshold=0.01); summary_table <- obj$qc_summary()

Evaluation signals

• qc_summary() output shows reduced feature count post-filter_group(); inspect 'filter' column for 'group' entries confirming blank-contaminated features were flagged.
• Manual inspection of a subset of removed features confirms they exhibit high intensity in blank samples and near-zero or minimal intensity in biological samples.
• Relative abundance threshold was applied consistently: verify that no feature in the blank group exceeds the chosen threshold (e.g., 0.01 or 1%) relative to non-blank samples.
• Re-running filter_group() on already-filtered data produces no new removals (idempotency check).
• Downstream statistical tests (fold change, t-tests) show reduced false-positive candidates compared to unfiltered data, indicating removal of non-biological signals.

Limitations

• Threshold selection (relative_abundance_threshold) is project-specific and depends on instrumental background, sample handling, and carryover rate; no universal default exists in the literature.
• filter_group() assumes that blank and biological samples are well-separated in intensity; features with intermediate blank–sample ratios may be ambiguous and prone to misclassification.
• If biological samples genuinely contain solvent or media (e.g., cells cultured in the same medium as blank), legitimate features may be incorrectly removed; domain knowledge of sample preparation is required to set appropriate thresholds.
• Multiple blank types (solvent, media, buffer) require separate filter_group() calls; cumulative removal may inadvertently eliminate true low-abundance metabolites if thresholds are not re-evaluated after each step.

Evidence

• [methods] filter_group() with group removal and relative abundance threshold: "Apply filter_group() to remove Solvent_Blank contaminant features above relative abundance threshold of 0.01. Apply filter_group() to remove Media blank features above relative abundance threshold of"
• [readme] Purpose of group filter in the package: "filter_group(): removal of features overrepresented in a specific group of samples; for example removal of features present in solvent blanks due to carryover between samples."
• [methods] qc_summary() output structure for interpreting filter results: "The qc_summary() function returns a data.table with compound IDs and filtering status for each ion, where passing ions are marked as passed and failing ions report the name of the filter they failed."
• [abstract] R6 reference semantics enabling in-place updates: "operates on reference semantics in which data is updated in-place. Compared to a shallow copy, where only data pointers are copied, or a deep copy, where the entire data object is copied in memory"