IP‑MS QC & Acceptance Criteria for IP Mass Spectrometry: LOD/LOQ, CV, Negative Controls, Background, Batch Effects

When you commission ip mass spectrometry for endogenous interactome work, the question isn't "did we run the pulldown?"—it's "do the acceptance criteria hold under scrutiny, are negative controls persuasive, and can the report be audited and reproduced?" This guide turns QC into contract-grade language you can paste into a SOW while staying reviewer‑friendly and fit‑for‑purpose.

Key takeaways

• Acceptance criteria should be measurable, traceable, and reproducible—no fixed universal promises; document definitions, methods, and limitations.
• Negative controls (Input, IgG, beads‑only) are non‑negotiable; KO/KD or orthogonal negatives raise confidence when feasible.
• Sensitivity is reported as LOD/LOQ with clear estimation methods (e.g., spike‑ins, dilution series), not fixed numbers.
• Reproducibility spans run, prep, and biology; summarize CV and agreement transparently and separate batch effects from biology.
• Background is quantified using in‑study negatives and contaminant resources (e.g., CRAPome), with effect‑size + FDR framing.
• Provide batch transparency: design, detection plots (e.g., PCA), mitigation, and decision rules.

Why QC is the real differentiator in ip mass spectrometry (and what "acceptance criteria" means)

Procurement and reviewers share the same core demand: data that withstand challenge. In IP‑MS, acceptance criteria should make it obvious what was measured, how sensitivity and specificity were supported, how reproducibility was demonstrated, how background was controlled, and how batches were tracked and mitigated. Practically, acceptance criteria are statements that are:

• Measurable (defined metrics and figures)
• Traceable (method, assumptions, provenance)
• Reproducible (replicate layers and batch handling visible)

If these elements are explicit, the same package serves audit, manuscript, and SOW/SLA needs without translation.

Two audiences, one QC language: reviewers vs procurement

• Reviewers care about scientific defensibility: specificity (controls, scoring), effect size with false discovery control, and whether claims match evidence.
• Procurement needs contract‑grade clarity: which controls were run in parallel, what must show agreement, how sensitivity was estimated, which plots/tables will appear, and what triggers rework.

The solution: one QC language that names controls, defines estimation/analysis methods, lists deliverables, and states decision rules with caveats.

Before you talk numbers: define the measurement target (what exactly is being quantified?)

Acceptance starts by naming the claim. In endogenous Co‑IP‑MS, your primary claims generally fall into one of four categories, each implying different QC emphasis:

• Presence/identification (is the bait present; are interactors detected beyond background?)
• Enrichment/interactome support (are putative interactors significantly enriched versus negatives?)
• Complex shift/mechanism of action (MoA) under perturbation (do interactions change under KO/KD, drug, or condition?)
• Absolute or anchored quantification (how much of a given protein/complex component is present?)

Claim‑to‑QC mapping (table‑ready)

A practical definition of "fit‑for‑purpose" QC

"Fit‑for‑purpose" means QC tightens or relaxes based on intended use: rebuttal to reviewers, early MoA exploration, or decision‑grade package for submission. Specify the intended use in the SOW and shape controls, sensitivity estimation, and batch handling accordingly. For deeper planning guidance, see the internal resource on the IP‑MS workflow design: controls, replicates, and planning.

Negative controls: the backbone of IP‑MS QC

Negative controls falsify alternative explanations. For endogenous Co‑IP‑MS, three controls are "must‑have," while stronger options further raise confidence when feasible.

Core controls that are universally defensible (Input + IgG + beads‑only)

SOW‑style clauses you can paste and adapt:

• "Process Input (matched total lysate), IgG isotype control, and beads‑only control in parallel with samples, using identical buffers, wash stringency, and instrument settings. Include them in all analysis steps and visualizations."
• "Report enrichment metrics comparing bait IP to IgG and beads‑only, and contextualize with Input abundance."
• "Document any proteins removed due to non‑specificity, with rationale (e.g., high frequency in controls or contaminant repositories). Provide a removed‑list appendix."

Rationale and evidence:

• Matched IgG and beads‑only reveal non‑specific antibody and matrix/solid‑support binders; Input confirms bait presence and scales enrichment. Contaminant repositories such as the CRAPome remain standard references for frequent background proteins in AP/IP‑MS pipelines, providing community‑curated context for filtering and interpretation according to the CRAPome original report in Nature Methods (2013) and subsequent reviews discussing its use.

Strong controls that raise confidence (KO/KD or orthogonal negative)

• "Where feasible, include a KO/KD lysate for the bait (or an orthogonal negative antibody) processed in parallel. Use it to demonstrate loss of specific interactors and to calibrate probabilistic or comparative scoring thresholds."
• "If KO/KD is infeasible, justify the choice of orthogonal negatives and disclose limitations."

These strengthen causal interpretation in endogenous systems and sharpen discrimination in scoring frameworks such as SAINT/CompPASS/MiST, which expect calibrated negatives.

How controls should appear in the report

• Side‑by‑side enrichment tables (bait vs IgG and beads‑only), volcano plots highlighting interactors that pass FDR and effect‑size criteria, Input‑anchored context for abundance, and an explicit "removed due to background" appendix.
• State how background reference information (e.g., CRAPome frequency) was consulted and how it influenced decisions. For an analysis‑methods refresher (filtering, normalization, FDR), see the IP‑MS data analysis workflow (filtering, normalization, FDR). For wet‑lab execution specifics and common failure modes in endogenous setups, see the Endogenous Co‑IP‑MS protocol checklist and failure modes.

Peer‑reviewed protocol notes emphasize matched controls and rigorous wash/handling as central to Co‑IP reproducibility; see STAR Protocols guidance on optimized Co‑IP execution with practical QC checkpoints in the Lagundžin et al. protocol (2022).

LOD/LOQ for IP‑MS: how to define, estimate, and report it transparently

Sensitivity is not a single number—it's a documented procedure. Define what "detectable" (LOD) and "quantifiable" (LOQ) mean for your matrix, platform, and targets, then show how you estimated them, with caveats.

LOD vs LOQ in plain terms

• LOD: the lowest level distinguishable from blank/background at a stated confidence.
• LOQ: the lowest level at which predefined precision/accuracy are achieved for quantification.

In proteomics, ion suppression, digestion efficiency, and missingness make universal thresholds unrealistic. Report the model and evidence, not just the outcome.

Fit‑for‑purpose ways to estimate sensitivity

• Matrix‑matched spike‑ins: Add known quantities of peptides/proteins into your lysate to measure recovery, linearity, and extrapolate LOD/LOQ with regression‑based rules under stated assumptions.
• Serial dilutions: Build calibration/dilution curves in the actual matrix to define linear range; compute LOD/LOQ using accepted analytical criteria and show residual diagnostics.
• Target‑specific verification: For critical proteins/interactors, perform replicate‑level checks focused on those targets to confirm quantifiability under your acquisition method.

Context and reviews on sensitivity/dynamic range and targeted quant workflows: see the comprehensive overview of bottom‑up proteomics sensitivity and dynamic range in Jiang et al., ACS Meas Sci Au (2024) and a targeted‑quant exemplar discussing absolute protein quantification and LOD/LOQ context in Williams et al., ACS Omega (2023).

What reviewers/procurement expect to see

• A short "Sensitivity Methods" paragraph naming the estimation approach (spike‑in, dilution), the statistical rule used, which figures/tables support it, and limitations (matrix‑specific suppression, carryover risk, peptide uniqueness, proteotypic coverage).
• Where absolute or anchored quant is part of the claim, clarify calibration materials, curve fit quality, and how uncertainty propagates to decision rules. Avoid fixed promises; emphasize verification within the stated conditions.

Reproducibility: CV, replicate agreement, and batch transparency

Reproducibility only makes sense when you separate replicate layers and disclose how batches were handled.

What reproducibility means in IP‑MS (run, prep, biology)

• Run‑level (instrument): repeated injections of the same prep to assess instrumental stability.
• Prep‑level (sample processing): independent preparations from the same biological material to capture processing variability.
• Biological level: independent biological sources/conditions capturing inherent biology.

State which layer each CV or correlation summarizes, and avoid conflating them.

CV and agreement metrics: what to report

• Per‑protein CV summaries for bait and key interactors within the appropriate replicate layer; distribution plots or heatmaps give context.
• Agreement metrics for replicate profiles (Pearson/Spearman for LFQ intensities or spectral counts), with missingness handling disclosed.
• A short text on imputation strategy (if used) and sensitivity analyses that show robustness of conclusions to missing data.

Batch effects: how to detect and document

• Design: randomize sample order; if multiple sessions are inevitable, include bridging/overlap samples and document lot/instrument/date.
• Detection: show PCA/MDS and variance decomposition pre/post normalization to demonstrate that batch structure is controlled relative to biology.
• Mitigation: use appropriate normalization (median, quantile, or model‑based) and, if necessary, rerun or reprocess outlier batches. Document all decisions.

For practical principles on batch correction and omics harmonization that translate to proteomics packages, see Dammer et al., Frontiers in Systems Biology (2023).

Background assessment: sticky proteins, non‑specific binding, and "signal‑to‑background"

Background isn't a vibe—it's a measurable set. Define it from your negatives and reference resources, then judge enrichment against it.

A practical definition of background in IP‑MS

Background is the set of proteins frequently or strongly observed in negative controls (IgG, beads‑only), optionally cross‑referenced with contaminant catalogs such as the CRAPome (Nature Methods, 2013). Treat it as experiment‑ and matrix‑specific; document how you built and used it.

Background acceptance criteria (table‑ready)

For context on AP/IP‑MS interaction scoring practices and interpretation (including CRAPome usage alongside probabilistic or comparative scores), see review discussions such as Kattan et al., Expert Review of Proteomics (2023).

How to present background in the report

• Include volcano plots that clearly mark negatives and accepted interactors; provide a concise FDR/effect‑size legend readers can audit.
• Summarize how contaminant frequency informed filtering or de‑prioritization. For plot interpretation primers and deeper dives, see the optional resources on the volcano plot for IP‑MS and on sticky proteins/background in IP‑MS. Analytical workflow details belong in the IP‑MS data analysis workflow (filtering, normalization, FDR).

The acceptance criteria template (SOW‑ready): what gets accepted, what triggers rework

IP‑MS QC and acceptance criteria framework covering negative controls, LOD/LOQ definition, reproducibility, background assessment, and batch effects.

Use this template to standardize expectations. Adapt wording to your bait, matrix, and intended use.

Must‑have acceptance criteria

• Controls and design
  • Input, IgG, and beads‑only processed in parallel; identical buffers/washes and acquisition settings; included in all analyses.
  • Document sample randomization; record batch‑defining covariates (instrument, date, lot, operator).
• Sensitivity and specificity
  • State LOD/LOQ definitions and the estimation approach (e.g., matrix‑matched spike‑ins or dilution series), with figures and uncertainty.
  • Use an interaction confidence framework (e.g., SAINT/SAINTexpress, CompPASS, or MiST) with dataset‑specific thresholding and disclosed assumptions.
• Reproducibility and agreement
  • Summarize CVs for bait and key interactors by replicate layer (run/prep/biology); provide replicate correlation plots and missingness handling.
• Background and filtering
  • Define an in‑study background from negatives; optionally annotate with CRAPome frequency; provide a "removed due to background" appendix.
• Reporting and traceability
  • Provide a methods summary naming software versions and parameters; include figure/table numbers that substantiate each acceptance point; ensure raw files and processed tables are traceable.

Should‑have (gold standard) criteria

• Strong negatives
  • KO/KD or orthogonal negative processed in parallel (if feasible) to strengthen causal interpretation and calibrate thresholds.
• Quantification quality
  • Calibration/linearity diagnostics and residual checks for any absolute/anchored claims; uncertainty propagation to decisions.
• Batch transparency
  • PCA/MDS and variance decomposition pre/post normalization; rationale for chosen normalization; bridging sample performance if multi‑batch.
• Complete visuals
  • Volcano plots with clear effect‑size and FDR marks; enrichment tables for bait vs each negative; CV distribution plots.

Rework triggers and decision rules

• Negative controls fail to differentiate from bait IP (e.g., IgG profile indistinguishable from bait within stated metrics) without a documented experimental explanation.
• Sensitivity reporting lacks a defined estimation method, figures, or uncertainty—no LOD/LOQ definition in context.
• Reproducibility collapses (e.g., discordant replicate layer agreement) without disclosed causes or remedial action.
• Batch structure dominates biology post‑mitigation, with no path to correction or re‑acquisition.
• Background handling is opaque (no removed‑list, no stated criteria) or contradicts known contaminant patterns without justification.

If you want these acceptance criteria tailored to your target, matrix, and timeline, talk with our team for a QC‑aligned IP‑MS plan via the IP‑MS absolute quantification service guide (deliverables, NDA/IP, fast quote).

How to align QC expectations before you start (avoid surprises)

Agreement upfront saves cycles later. Use this kickoff checklist as prose you can paste into a scope email.

• Name the study claim and intended use (manuscript rebuttal, MoA exploration, decision‑grade package, etc.).
• Confirm bait presence/abundance, available negatives (Input, IgG, beads‑only), and feasibility of KO/KD or orthogonal negatives.
• Set expectations for sensitivity reporting (which estimation method, which targets) and list the specific figures/tables you expect.
• Define replicate layers and sample counts (run/prep/biology), plus randomization and any bridging samples.
• List batch‑defining covariates you will record; agree on normalization approach classes and plots to include.
• Clarify background handling and the criteria for removal vs de‑prioritization, with an explicit appendix in the final report.
• State deliverables format and access (raw files, processed tables, figure bundle), and align on NDA/IP.

What to send us for a QC‑aligned quote

• Sample type and count; bait identity and expected abundance.
• Claim category (presence, enrichment, MoA shift, absolute/anchored).
• Available negatives (Input/IgG/beads‑only) and feasibility of KO/KD or orthogonal negatives.
• Critical targets for sensitivity verification; required timelines and milestones.
• Compliance constraints (NDA/IP, data security) and any formatting requirements for submission.

For timeline, deliverables, and confidentiality alignment, see the IP‑MS absolute quantification service guide (deliverables, NDA/IP, fast quote).

Notes on interaction scoring frameworks (for acceptance language)

Probabilistic or comparative scoring translates enrichment evidence into interpretable confidence with calibrated error control. Your SOW should name the chosen framework and state how thresholds were derived for this dataset.

• SAINT/SAINTexpress (probability‑based; reports AvgP): foundational description in Nature Methods (2011); streamlined implementation described in Teo et al., J Proteomics (2014). Label‑free intensity‑based scoring is covered in SAINT‑MS1, J Proteome Res (2012).
• CompPASS (comparative ranking and specificity weighting) pioneered large‑scale AP‑MS prioritization; see the original Cell (2009) report and methodological overviews in later methods papers.
• MiST (specificity‑driven confidence, 0–1) complements these approaches; see the Proteomics methods paper in Trigg et al., Proteomics (2012).

When you include these scores in acceptance criteria, add a sentence like: "Thresholds were calibrated to achieve a study‑level false discovery target using in‑study negatives and known complex members; limitations due to missingness and bait coverage are disclosed."

Example acceptance language blocks (paste‑ready)

Use these as starting points and customize to your scope.

• Controls clause: "We will process Input, IgG, and beads‑only controls in parallel with identical buffers, washes, and acquisition settings. All controls will be included in analysis, visualized alongside samples, and used to define the in‑study background."
• Sensitivity clause: "Sensitivity (LOD/LOQ) will be defined for the stated matrix and targets using [matrix‑matched spike‑ins | serial dilutions], with regression diagnostics, uncertainty intervals, and figure/table references reported. No fixed sensitivity levels are promised beyond the validated operating range."
• Reproducibility clause: "Run, preparation, and biological replicates will be distinguished. We will report per‑protein CVs for the bait and key interactors within the appropriate layer, replicate agreement metrics, and missingness handling."
• Background clause: "We will construct an in‑study background from IgG and beads‑only, optionally informed by community contaminant frequency resources, and provide a ‘removed due to background' appendix with justification."
• Batch clause: "We will randomize sample processing/order, record batch‑defining covariates, display PCA/MDS and variance decomposition pre/post normalization, and document any corrective actions."
• Scoring clause: "Interaction confidence will be summarized using [SAINT/SAINTexpress | CompPASS | MiST] with dataset‑specific thresholding and disclosed assumptions/FDR calibration."

Frequently raised objections (and how transparent QC addresses them)

• "We can't accept the data without KO/KD." When KO/KD is infeasible, document why and reinforce with orthogonal negatives and strong IgG/beads‑only contrasts; disclose limitations in the acceptance text.
• "These CVs look high." Specify the replicate layer, show distributions and agreement plots, and discuss whether uncertainty changes the decision. High CV at the run layer suggests instrument issues; at the biology layer it may reflect real heterogeneity.
• "Background seems too high." Show the definition and size of the in‑study background, effect sizes against it, and the removed‑list with rationale. Reference contaminant frequencies where relevant rather than asserting "low/high."
• "Batch effects dominate." Present design choices (randomization, bridging), detection plots, and mitigation results. If residual batch remains, propose defined rework actions.

Next step

Share your claim type, available controls, and sample constraints—we'll propose a QC plan and reporting package aligned with your manuscript or SOW. See the IP‑MS absolute quantification service guide (deliverables, NDA/IP, fast quote).

Author

CAIMEI LI
Senior Scientist at Creative Proteomics
LinkedIn: https://www.linkedin.com/in/caimei-li-42843b88/

I specialize in IP‑MS study design and reporting, with a focus on QC transparency, quantitative reproducibility, and deliverables that pass reviewer and procurement audits.

Disclaimer: For research purposes only, not intended for clinical diagnosis, treatment, or individual health assessments.

References (peer‑reviewed; journal/DOI pages)

• Mellacheruvu et al., The CRAPome: a contaminant repository for affinity purification–mass spectrometry data. Nature Methods (2013). https://doi.org/10.1038/nmeth.2556
• Jiang et al., Comprehensive Overview of Bottom‑Up Proteomics Using LC–MS. ACS Meas Sci Au (2024). https://doi.org/10.1021/acsmeasuresciau.3c00068
• Williams et al., Discovery Proteomics and Absolute Protein Quantification. ACS Omega (2023). https://doi.org/10.1021/acsomega.2c07614
• Choi et al., Probabilistic scoring of affinity purification–mass spectrometry data (SAINT). Nature Methods (2011). https://www.nature.com/articles/nmeth.1546
• Teo et al., SAINTexpress: Improvements and simplification for interaction scoring. J Proteomics (2014). https://doi.org/10.1016/j.jprot.2014.01.023
• Trigg et al., A novel methodology for interactome confidence scoring (MiST). Proteomics (2012). https://doi.org/10.1002/pmic.201100531
• Sowa et al., Defining interactomes with comparative scoring (CompPASS). Cell (2009). https://doi.org/10.1016/j.cell.2009.07.033
• Dammer et al., Batch correction and harmonization of –omics datasets. Frontiers in Systems Biology (2023). https://doi.org/10.3389/fsysb.2023.1092341
• Kattan et al., Analysis of affinity purification‑related proteomic data for interactome mapping. Expert Review of Proteomics (2023). https://doi.org/10.1080/14789450.2023.2176307