Proteomics

Mass spectrometry-based proteomics data analysis package for the Datagrok platform.

It takes a study from raw search-engine output to a result a biologist can act on — which proteins changed, what pathways they implicate, and what each protein is — and lets you share that result as a read-only snapshot a collaborator can explore without touching the raw data or writing a line of code. It covers label-free quantitative proteomics from both DDA (MaxQuant, FragPipe) and DIA (Spectronaut) workflows.

Open it from the Proteomics top menu.

One screen from statistics to biology: the interactive volcano, the clustered expression heatmap, and in-place UniProt annotation for the selected protein.

Where it fits

The package picks up where the search engine leaves off and carries the study all the way to biological interpretation. It does not replace MaxQuant, FragPipe, DIA-NN or Spectronaut, and it does not touch raw instrument files (.raw, .mzML, .d) — you bring in the protein/quant table they produce. From there it stands in for the four-to-six disconnected desktop tools and hand-exported files that normally sit between that table and a finished figure: import, QC, statistics, interpretation, and sharing happen in one place, and the result stays reproducible instead of scattered across scripts and spreadsheets.

What you can achieve

If you generate the data — the proteomics scientist

• Stand behind your quantification. Normalization removes between-sample technical bias and principled missing-value handling deals with the dropout that is intrinsic to MS, so downstream fold-changes reflect biology rather than loading or detection artifacts.
• Make a defensible significance call. Statistics built for the small-sample, high-dimensional reality of proteomics (limma / DEqMS) give you a ranked, FDR-controlled list of changed proteins you can justify to a reviewer — not an ad-hoc cutoff.
• Catch problems before they become false discoveries. Per-run QC and longitudinal process control surface a mislabeled channel, a weak replicate, or instrument drift across a batch before it drives a result.
• Hand off something reproducible. Publish the finished analysis as a versioned, read-only snapshot — the collaborator sees exactly what you saw, and a later re-analysis supersedes rather than overwrites it.

If you interpret the data — the biology scientist

• See what changed. An interactive volcano shows which proteins go up or down between your conditions and how strongly — sorted, labeled, and clickable, not a static figure.
• Understand what it means. Enrichment collapses a list of hundreds of changed proteins into the handful of biological processes, pathways, and complexes that are actually over-represented (GO, KEGG, Reactome, WikiPathways) — and selecting a pathway highlights its proteins back on the volcano, so statistics and biology stay connected.
• Know what each hit is. Click any protein to pull its name, gene, function, and subcellular location from UniProt, in place — no separate database lookups.
• Explore it yourself. Open the shared snapshot and filter, select, and cross-link interactively. You don't need the raw files, the analysis environment, or another round trip through the analyst to ask your own follow-up questions.

Supported inputs

| Format | File | Notes | |--------|------|-------| | MaxQuant | proteinGroups.txt | DDA. LFQ / Razor / iBAQ intensities; contaminant & reverse-hit filtering | | Spectronaut — Report | PG-level export | DIA. Pivots long-form precursor exports to a protein × sample matrix; streams multi-GB files; q-value filtering | | Spectronaut — Candidates | pre-computed DE export | DIA. Already-computed fold-change / FDR per protein; skips straight to visualization | | FragPipe | combined_protein.tsv | DDA. MaxLFQ / Intensity / Razor; Philosopher contaminant filtering | | Generic matrix | any CSV/TSV | Column-mapping dialog; auto-detects delimiter and log2 status |

DIA-NN is not currently supported as a direct importer — bring DIA-NN output in via the Generic matrix path, or as a Spectronaut-style matrix.

Workflow

The pipeline runs as a sequence of steps from the Proteomics menu. Each step records its completion on the table, and downstream steps read those as preconditions, so the menu greys out actions that don't yet apply (and items that need per-sample intensities are disabled on the pre-computed Spectronaut Candidates table).

Import

Spectronaut Candidates... · Spectronaut Report... · MaxQuant... · FragPipe... · Generic Matrix... — also available from the left-bar main menu with no table open.

Annotate

• Annotate Experiment — assign intensity columns to two named groups (e.g. Control vs. Treatment); required input for normalization, imputation, DE, PCA and enrichment.

Analyze

• Normalize — median centering, quantile normalization, or VSN (variance-stabilizing; R, with a quantile fallback).
• Impute Missing Values — k-nearest-neighbor, MinProb (Perseus-style downshifted draw), mean, median, or zero.
• Differential Expression — DEqMS (peptide-count-weighted variance) → limma (empirical-Bayes moderated t-test) → client-side Welch's t-test, cascading down on availability. Adds log2FC, p-value, adj.p-value and a significant flag.
• Enrichment Analysis — over-representation against GO (BP/MF/CC), KEGG, Reactome and WikiPathways through g:Profiler, for 9 model organisms (human, mouse, rat, yeast, E. coli, zebrafish, fly, Arabidopsis, C. elegans), with optional smart GO pruning.
• Compute SPC Status — score the active run against an instrument-QC baseline.

Visualize

• Volcano Plot — log2FC vs. -log10(p), with significance threshold lines; switch the significance metric, color by significance or UniProt subcellular location, and label the top-N proteins independently of selection.
• Heatmap — clustered expression heatmap of the top differential proteins.
• PCA — sample-level principal-component scatter (opens in its own view).
• Group-Mean Correlation — group1 vs. group2 mean per protein, with Pearson/Spearman.
• QC Dashboard — MA plot, coefficient-of-variation, missingness matrix and an intensity trend.
• SPC Dashboard — longitudinal control charts over instrument metrics (median intensity, missingness, replicate correlation, protein count) with Nelson-rule flagging and a Pareto of tripped rules.
• Enrichment Charts — dot plot and bar chart of the top terms, split Up/Down by direction; selecting a term highlights its proteins back in the volcano.

Selecting a term in the enrichment results highlights that pathway's member proteins back on the volcano — a pathway you find in the biology view is immediately located in the statistics view, so the two never drift apart:

Selecting the KEGG Cell Cycle term (top) highlights its 25 member proteins on the volcano (bottom).

Share

• Share Analysis for Review — publish a trimmed, read-only snapshot (table + volcano + enrichment charts) under a project (chosen from an admin-curated list) to a reviewer team. The source table is never mutated; reviewer access is verified (and rolled back on an over-permissive grant); the snapshot is optionally re-opened and shape-checked to confirm it survives a reload; re-publishing supersedes the prior version. A Shared Analysis context panel surfaces the publication's audit trail.

Protein annotation

Clicking a protein-ID cell opens the UniProt info panel, which fetches the protein name, gene, organism, functional description, GO terms (MF / BP / CC) and KEGG / Reactome / InterPro cross-references from the UniProt REST API and caches them in App Data.

Semantic type detection

Columns are auto-detected and tagged so the package keeps working on tables from vendors it hasn't seen — detectors match on column-name hints and validate sample values.

| Semantic type | Detects | |---------------|---------| | Proteomics-ProteinId | Protein IDs, Majority Protein IDs, UniProt, Accession (validated against the UniProt accession pattern) | | Proteomics-GeneSymbol | Gene Names, Gene Symbol, Gene | | Proteomics-Log2FC | columns naming "log2" with "fc"/"fold" | | Proteomics-PValue | p-value, adj.p, FDR, q-value (range 0–1) | | Proteomics-Intensity | Intensity, LFQ / Reporter / Razor / MaxLFQ Intensity, iBAQ | | Proteomics-SubcellularLocation | UniProt subcellular-location strings | | Proteomics-DisplayName, Proteomics-SourceId | gene-label resolver fields | | Proteomics-NumeratorMean, Proteomics-DenominatorMean | per-group means in Spectronaut Candidates exports |

Integrations & requirements

• No credentials required. The package calls only public REST APIs — UniProt (protein annotation) and g:Profiler (enrichment).
• R is optional. limma_de.R, deqms_de.R and vsn_normalize.R run server-side and require a configured Datagrok R compute environment; every R path has a client-side TypeScript fallback, so the package is fully functional without R.

Configuration

Package administrators set these in Plugins → Proteomics → Settings (a custom editor replaces the default form). They gate and shape the Share flow:

| Setting | What it does | |---------|--------------| | Project vocabulary | Controlled list of project names an analyst can share under. Required before anyone can share — with none configured, the Share dialog blocks and asks for an admin. Add / rename / remove entries as editable rows. | | Default reviewer team | Team pre-selected in the Share dialog's team picker; blank falls back to the first available team. | | Review space name | Umbrella Space shared snapshots publish under; each project gets a child Space inside it. | | Review name prefix | Prefix for published project names (<prefix>-<project>-v<n>-<date>); the suffix is fixed so re-share versioning keeps working. | | Verify published dashboard | Re-open each snapshot after publishing to confirm it survives a reload (safer; off = faster demo shares). |

Demo data

Example datasets ship in files/demo/ (see its README.md):

• proteinGroups.txt — small MaxQuant example
• cptac-spike-in.txt — CPTAC UPS spike-in reference with known concentrations
• fragpipe-smoke-test.tsv — FragPipe combined_protein.tsv example
• spectronaut-hye-candidates.tsv — Spectronaut Candidates (pre-computed DE)
• spectronaut-hye-precursor.tsv / spectronaut-hye-mix.tsv — Spectronaut long-form reports

Development

cd packages/Proteomics

npm run build              # grok api && grok check --soft && webpack
grok publish local         # build + publish to your configured server
grok link && npm install   # link local datagrok-api / @datagrok-libraries
grok test --host localhost # run the package tests

See also

• Personas & capabilities — who (analyst vs biology scientist) can do what, and how the boundary is enforced.
• TODO: Datagrok Proteomics documentation — the help page at https://datagrok.ai/help/domains/bio/proteomics does not exist yet and still needs to be written.