# Performing Proteomics Quality Control

> Perform quality control for protein-level proteomics data. Use when filtering outlier samples or low-evidence features before downstream analysis. Covers PSM artifact removal, sample QC, feature-level filtering. Does NOT cover normalization, batch correction, or imputation.

- **Type:** Skill
- **Install:** `agentstack add skill-mannlabs-proteomics-agent-skills-performing-proteomics-quality-control`
- **Verified:** Yes — security-reviewed for prompt injection and unsafe behavior
- **Seller:** [MannLabs](https://agentstack.voostack.com/s/mannlabs)
- **Installs:** 0
- **Category:** [Agent Skills](https://agentstack.voostack.com/c/agent-skills)
- **Latest version:** 0.1.0
- **License:** Apache-2.0
- **Upstream author:** [MannLabs](https://github.com/MannLabs)
- **Source:** https://github.com/MannLabs/proteomics-agent-skills/tree/main/plugins/proteomics/skills/performing_proteomics_quality_control

## Install

```sh
agentstack add skill-mannlabs-proteomics-agent-skills-performing-proteomics-quality-control
```

Requires the [AgentStack CLI](https://agentstack.voostack.com/docs/cli). Works with Claude Code, Cursor, and any MCP-compatible agent.

## About

# Performing Proteomics Quality Control

Goal: Remove outlier samples and unsupported features from protein-level quantification matrices.

## Context and Definitions

### Key Metrics

- **Decoy prefixes**: `REV_`, `DECOY_`, or boolean decoy indicator column
- **Quality Control (QC) for technical replicates**: Remove technical replicates when coefficient of Variation (CV) > 20-30% or Pearson Correlation R  N × MAD(X)`, where _N_ is a user-defined factor. MAD is calculated as the median of absolute deviations from the median: `MAD(X) = median(|X - median(X)|)`.
- **Feature Completeness**: Feature-wise fraction of samples with non-missing values.

### Study Types

| Type         | Min features/sample | Completeness | Intensity metric |
| ------------ | ------------------- | ------------ | ---------------- |
| Single-cell  | >500-600 proteins   | 10-15%       | Total intensity  |
| Bulk tissue  | No strict minimum   | 50-70%       | Total intensity  |
| Plasma/serum | No strict minimum   | 20-50%       | Median intensity |

### Defaults

| Parameter                 | Default                                      | Adjust when                                                                                                                                                                                                          |
| ------------------------- | -------------------------------------------- | -------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- |
| False discovery threshold | 0.01                                         | Standard value                                                                                                                                                                                                       |
| MAD multiplier (N)        | 3                                            | Lower (3) more permissive                                                                                                                                                                       |
| Min unique peptides       | 2                                            | >2 for validation studies                                                                                                                                                                                            |
| Feature Completeness      | Study-dependent, typically between 10% - 70% | Must find a compromise between robustness and expected biological prevalence. If an effect is expected in a small subset of samples (e.g. a rare cell type) it should not be removed by feature completeness filters |

### Failed Sample Detection

**Removing failed samples is critical.** Samples with very low protein counts or intensity typically result from missed injections, sample degradation, or other laboratory errors. Since proteomics relies on label-free quantification (relative quantities across samples), failed samples corrupt normalization and distort all downstream results.

**Example**: In a plasma study, one sample has 30 protein groups while all others have ~500 protein groups—this outlier should be removed before normalization.

## Checklist

1. [ ] Clarify Parameters
2. [ ] Remove Low Confident Identifications
3. [ ] Remove Peptide Spectrum Match Artifacts
4. [ ] If technical replicates exist: Assess technical replicate reproducibility
5. [ ] Remove sample outliers
6. [ ] Filter features below completeness threshold
7. [ ] Generate QC report

**Order matters**: Steps 2-3 (PSM) → Steps 4-6 (Sample QC) → Step 7 (Feature QC). Sample QC must precede feature QC because outlier samples inflate apparent missingness.

## Instructions

### Step 1: Clarify Parameters

Before starting, check for:

- **Study type**: Single-cell, bulk, or plasma?
- **Completeness threshold**: What minimum biological prevalence is expected? (e.g., 10% cell population → expect 90% missingness for its markers)
- **Outlier handling**: Remove samples or flag only?
- **Technical replicates**: Present in dataset?

### Step 2: Remove Low Confident Identifications

- Remove rows with decoy prefixes or decoy indicator = True
- If applicable: Remove proteins with = false discovery threshold

### Step 4: Technical Replicate QC (if applicable)

- Compute CV and Pearson R between replicates
- Remove replicate sets with CV > 30% or Pearson R < 0.9

### Step 5: Sample QC

Compute per-sample:

- **Intensity**: Total (or median for plasma/serum)
- **Feature count**: Number of non-missing proteins

For each metric, remove MAD outliers (see [Key Metrics](#key-metrics))

Always perform analysis and remove outliers before proceeding.

### Step 6: Feature QC

- Compute completeness per feature
- Remove features below completeness threshold

### Step 7: Generate QC Report

Output:

1. **Summary table**: That shows for each quality control step the number of samples and features before and after filtering. For an example see ./references/qc-summary.tsv
2. **Histogram of each metric**: Distribution of the respective metric before filtering on the left, and after filtering on the right.
3. **Filtered matrix**: With observations as rows and features as columns. Ready for downstream processing

## Source & license

This open-source skill is cataloged on AgentStack and links to its original source — we do not rehost the code.

- **Author:** [MannLabs](https://github.com/MannLabs)
- **Source:** [MannLabs/proteomics-agent-skills](https://github.com/MannLabs/proteomics-agent-skills)
- **License:** Apache-2.0

Install and usage instructions live in the source repository linked above.

## Pricing

- **Free** — Free

## Security capabilities

Automated source analysis of v0.1.0 — what this tool can access:

- **Network access:** no
- **Filesystem access:** no
- **Shell / process execution:** no
- **Environment & secrets:** no
- **Dynamic code execution:** no

*"Yes" means the capability is present in the source — more access means more to trust, not that it is unsafe.*


## Versions

- **0.1.0** — security scan: passed — Imported from the upstream source.

## Links

- Listing page: https://agentstack.voostack.com/l/skill-mannlabs-proteomics-agent-skills-performing-proteomics-quality-control
- Seller: https://agentstack.voostack.com/s/mannlabs
- Browse the marketplace: https://agentstack.voostack.com/browse

---
Listed on AgentStack — the marketplace for AI agent skills and MCP servers. Every listing is security-reviewed. Creators keep 70%.