Proteomics

Single-cell proteomics analysis: carrier-multiplexed MS quantification from PSM tables to annotated cell-state clusters

Single-cell proteomics quantifies proteins per cell to resolve cell-state heterogeneity that bulk LC-MS/MS averages away. Pepkio runs version-pinned analysis from PSM or protein matrices to cell-state clusters and differential-abundance tables. Custom work is scoped at kickoff. For academic, biotech, and pharma clients. SCoPE2 reports about 1,000 proteins per cell [1]. R scripts, figures, and a Methods draft are included.

Key facts

Key facts about Single-Cell Proteomics
FactValue
Supported platforms / instrumentsThermo Orbitrap (Exploris, Fusion Lumos, Eclipse); Bruker timsTOF (dia-PASEF, Ultra when scoped); TMT/TMTpro isobaric carrier multiplexing (SCoPE2-style; [1]); plexDIA and nPOP; label-free DIA single-cell; Proteome Discoverer, MaxQuant, and DIA-NN exports. Scoped at kickoff.
Input requirementsPSM or protein quantification tables plus colData with cell ID, MS batch/run, condition, and carrier channel; often hundreds of cells per condition for clustering [2]; diluted bulk reference sample per batch recommended [2]
Reference builds supportedUniProt Swiss-Prot reviewed proteomes (human UP000005640, mouse UP000000589); CPTAC-style contaminant sequences; organism-specific FASTA on request
Primary tools (with versions)scp 1.22+; QFeatures; SingleCellExperiment; QuantQC; DIA-NN 2.0+; MaxQuant 2.8+; proDA 1.0; limma 3.60, pinned per project
Typical turnaround time4-8 weeks (standard SCP cohort, one primary contrast). Longer for multi-batch integration, matched scRNA-seq, or atlas-scale cell counts. Confirmed at kickoff.
Deliverable formatsProtein matrices (.csv, .tsv); SingleCellExperiment .rds objects; PDF/SVG figures; HTML QC report; commented R scripts; Methods draft
Key cited best-practice referenceDeutsch et al. [2], Nature Methods; Specht et al. [1], Genome Biology (SCoPE2)
Custom / bespoke analysisNon-standard matrices, cross-modality integration, client-specified QC thresholds, trajectory analysis, or phosphoproteomics extensions scoped at kickoff

What is single-cell proteomics?

Single-cell proteomics (SCP) converts PSM tables into a cell-by-protein matrix, then applies per-cell QC, batch correction, clustering, and differential testing. That measures which proteins define cell states rather than tissue averages [2]. Unlike bulk DDA/DIA proteomics, SCP can resolve subpopulations and continuous proteome gradients [1]. Unlike Olink PEA, it is discovery-scale MS without a fixed antibody panel. SCoPE2 quantified about 1,000 proteins per cell across 1,490 monocytes and macrophages [1]. Pepkio starts from client PSM or protein matrices with documented QC. Custom workflows are agreed at kickoff. See the single-cell proteomics glossary.

When should you use single-cell proteomics?

Single-cell proteomics fits when you need protein-level cell-state heterogeneity, such as differentiation trajectories, rare subpopulations, or post-transcriptional regulation that bulk or RNA measurements often miss [1, 2].

Comparison of single-cell MS proteomics, bulk DDA/DIA, scRNA-seq, and Olink PEA approaches
ApproachBest forLimitationsApproximate cost range
Single-cell MS proteomics (SCoPE2 / DIA / nPOP)Cell-state discovery, proteome gradients, low-input heterogeneity, matched multi-omicsSparse matrices; carrier or library design affects depth; MS time per cell limits throughputHigher per-cell MS and sample-prep cost than bulk; moderate bioinformatics cost
Bulk DDA/DIA proteomicsCohort-scale differential abundance when tissue is homogeneousAverages across cell types; rare states below detectionLower per-sample MS cost than SCP for modest cohorts
scRNA-seqCell-type atlases, trajectory inference, large cell countsmRNA proxy; post-transcriptional regulation and protein stability are not measured directly ([1]: about 20-fold more protein than RNA copies per gene)Library prep + sequencing; bioinformatics varies by cell count
Olink PEATargeted plasma or CSF panels, clinical-scale validationFixed panel size; not discovery-scale MS depthPanel and plate cost; efficient for targeted biomarker testing
  • Macrophage differentiation: Specht et al. [1] profiled 1,490 monocytes and macrophages by SCoPE2. They observed a continuous proteome gradient during differentiation rather than discrete cytokine-polarization clusters.
  • ES cell subclusters: Rosenberger G et al. (2023) applied sensitivity-tailored DIA to mouse embryonic stem cells and identified proteome subclusters with distinct metabolic enzyme expression.
  • Hepatocyte zonation: Rosenberger FA et al. [3] used scDVP to map about 1,700 proteins per hepatocyte slice; half of the proteome was differentially regulated near the central vein.

How the analysis works — step by step

  1. 1. Validate inputs and experimental design

    Pepkio confirms PSM or protein matrix integrity, MS batch structure, carrier channels, and contrast definitions in sample_manifest.csv. Confounded batch designs or very few cells per group are flagged before processing [2].

    Tools and outputs

    Tools used: Custom validation scripts; scp annotation schema check

    Output: sample_manifest.csv with cell IDs, MS batch, condition labels, carrier design notes, and QC flags

  2. 2. Import PSM tables into QFeatures structure

    PSM quantification tables from MaxQuant, Proteome Discoverer, or DIA-NN are imported per MS batch using readSCP() into a QFeatures object with a PSM to peptide to protein hierarchy [4, 5, 6].

    Tools and outputs

    Tools used: scp 1.22+; QFeatures; SingleCellExperiment

    Output: Per-batch QFeatures object; import log with row and column counts per set

  3. 3. Run MS-batch and cell-level QC

    Run-level and per-cell metrics include reporter intensity (RI), CV, peptide counts, and sample-to-carrier ratio (SCR). QuantQC is applied when nPOP exports include it [2, 7].

    Tools and outputs

    Tools used: scp 1.22+; QuantQC; custom R diagnostics

    Output: cell_qc_summary.csv; SCR distribution plots; RI and peptide-count histograms per batch

  4. 4. Filter low-confidence PSMs

    PSMs are filtered at 1% peptide FDR. SCR-based filtering removes PSMs with unreliable single-cell quantification. Carrier-only PSMs (missing SCR) are excluded [2, 4].

    Tools and outputs

    Tools used: scp 1.22+

    Output: Filtered PSM assay; psm_filter_log.csv documenting FDR and SCR thresholds

  5. 5. Aggregate PSM → peptide → protein

    PSM intensities are aggregated to peptide and protein levels using median aggregation per scp conventions, preserving links between assay layers for audit and re-aggregation [4].

    Tools and outputs

    Tools used: scp 1.22+; QFeatures aggregation functions

    Output: Peptide and protein assays within QFeatures; aggregation_log.csv

  6. 6. Convert zeros to missing values and normalize

    Zero intensities are converted to NA before normalization, treating zeros as missing rather than true absence [2, 4]. Protein matrices are log-transformed. MNAR versus MCAR missing-value patterns are profiled before any imputation [8].

    Tools and outputs

    Tools used: scp 1.22+

    Output: protein_matrix_raw.csv; missing_value_profile.csv; normalization parameters in analysis_parameters.yaml

  7. 7. Filter low-quality cells

    Cells with low RI, high CV, or very few quantified peptides are removed using batch-adaptive thresholds. Retained and excluded counts are documented per filter rule [2].

    Tools and outputs

    Tools used: scp 1.22+

    Output: Filtered protein matrix; updated cell_qc_summary.csv; cell-exclusion plots

  8. 8. Correct batch effects across MS runs

    SCP acquires few multiplexed cells per MS run, so batch structure is typical. scplainer removes technical variation when shared cell states span batches. Batch is modeled as a covariate when correction is not possible [2].

    Tools and outputs

    Tools used: scp 1.22+ (scplainer); limma 3.60 when batch is modeled

    Output: Batch-corrected protein matrix; before/after PCA by batch and condition

  9. 9. Cluster, embed, and annotate cell states

    Pepkio builds a neighbor graph on the corrected protein matrix, runs Leiden community detection at data-driven resolution, and computes PCA/UMAP embeddings. Cell states receive provisional labels from top marker proteins per cluster.

    Tools and outputs

    Tools used: scp 1.22+; leidenalg 0.10+; custom R visualization

    Output: Cluster assignments in cell_metadata.csv; PCA/UMAP plots; marker_protein_table.csv

  10. 10. Test differential protein abundance and package deliverables

    Cluster-wise or condition-wise differential testing uses proDA or limma with Benjamini-Hochberg FDR control, accounting for missing values without blind imputation [9, 10]. Optional GO/KEGG enrichment is scoped at kickoff. Figures, scripts, README, and a Methods draft are bundled.

    Tools and outputs

    Tools used: proDA 1.0; limma 3.60

    Output: dep_by_cluster.csv or dep_<contrast>.csv; volcano plots; final .rds object; scripts; Methods draft

What Pepkio delivers

Processed data

  • Raw and filtered protein matrices (.csv, .tsv)
  • SingleCellExperiment .rds objects
  • cell_metadata.csv with QC metrics, cluster, and provisional cell-state labels
  • analysis_parameters.yaml documenting FDR, SCR, normalization, and batch-correction choices

Figures (PDF/SVG)

  • SCR and RI histograms
  • Per-cell peptide and protein count plots
  • PCA and UMAP (cluster, condition, MS batch)
  • Marker-protein heatmaps
  • Composition bar plots
  • DE volcano plots per contrast or cluster

Tables

  • sample_manifest.csv, cell_qc_summary.csv, psm_filter_log.csv, missing_value_profile.csv
  • marker_protein_table.csv
  • dep_by_cluster.csv or dep_<contrast>.csv with columns ProteinID, Gene, log2FC, SE, pvalue, adj.pvalue

Code

  • Commented R scripts per analysis stage
  • Environment lock files (sessionInfo(), renv lock, or conda export)
  • Delivery via private Git repository or agreed file transfer

Documentation

  • HTML QC report with thresholds and outlier flags
  • README with reproduction instructions
  • Methods draft citing pinned software versions and filtering parameters
  • Post-delivery reviewer support for clarification and minor revisions within agreed scope (typically ≤20% of project scope)

Technical decisions we make — and why

Data structure: QFeatures with one set per MS batch
Preserves SCR, carrier channel, and batch metadata that flat matrices discard [4]. Alternative: a single flat protein matrix, which loses per-batch QC and SCR filtering.
PSM filtering: 1% FDR plus SCR-based filtering
SCR removes PSMs where single-cell reporter signal is unreliable relative to the isobaric carrier [2]. Alternative: FDR-only filtering, which can retain peptides identified via the carrier but poorly quantified in single cells.
Missing values: convert zeros to NA; proDA for DE
SCP missingness is predominantly MNAR, and blind imputation can inflate false positives [4, 8]. Alternative: KNN or MinDet imputation, scoped when the design and missingness profile support it.
Batch correction: scplainer default; batch covariate in limma when not confounded
MS-run batch effects are common in SCP and can mimic biology if ignored [2]. Alternative: no correction when batch and condition are confounded, documented in the QC report.
Carrier-aware channel handling: exclude affected TMT channels at high carrier ratios when needed
Isotopic impurities from the carrier channel (126) affect channels 127N and 128C at high carrier ratios [11, 12]. Channel exclusion strategy is documented at kickoff.

Common questions

What is the minimum number of cells and proteins per cell for single-cell proteomics analysis?

Often hundreds of cells per condition support clustering. Deutsch et al. [2] recommend reporting counts per group, and Specht et al. [1] analyzed 1,490 cells. SCoPE2 reports about 1,000 proteins per cell [1]. nPOP with plexDIA reports about 3,000-3,700 proteins per human cell [7]. Smaller cohorts are accepted with reduced-power notes in the QC report. Exact targets are confirmed at kickoff.

Can you analyze low-yield or failed single cells?

Yes, with caveats documented in the QC report. Cells with very low reporter intensity, high CV, or very few quantified peptides are flagged and typically excluded before clustering [2]. Empty carrier channels and carrier-only PSMs are removed during SCR filtering. Re-acquisition or pooling strategies are discussed when cell yield threatens the study question.

Do you support Thermo Orbitrap, Bruker timsTOF, TMT multiplex, and plexDIA data?

Yes, for formats we can load after kickoff review. Thermo Orbitrap .raw files from Exploris, Fusion Lumos, and Eclipse support TMT/TMTpro carrier multiplexing and DIA workflows when scoped. Bruker timsTOF .d files support dia-PASEF and Ultra single-cell acquisitions when scoped. MaxQuant, Proteome Discoverer, and DIA-NN PSM or protein exports are the primary computational entry points. nPOP and plexDIA exports are supported when scoped at kickoff.

How long does single-cell proteomics analysis take at Pepkio?

A standard SCP project (roughly 100-500 cells, one primary contrast, one or two MS batch groups, and clustering plus DE) typically completes in 4-8 weeks from data receipt. Multi-batch integration, matched scRNA-seq correlation, or cohorts exceeding 1,000 cells may take longer. Milestone check-ins occur during the project. Exact timelines are confirmed at kickoff.

How do you handle batch effects across MS acquisition runs or prep dates?

When batch is known and not fully confounded with condition, Pepkio applies scplainer batch correction or includes batch as a covariate in limma/proDA models [2]. PCA and RI/CV diagnostics are reviewed before modeling. Designs without batch randomization are flagged at kickoff. Biological and technical factors may be inseparable when randomization was not performed.

Do I own the code, and in what format is it delivered?

Yes. You retain full ownership of all code, scripts, and results delivered under the project agreement. Pepkio provides commented R scripts with environment lock files so you can rerun analyses when the execution environment matches the pinned setup. Matrices use standard .csv and .tsv formats. SingleCellExperiment objects are delivered as .rds files readable in R/Bioconductor.

Can I be involved during analysis?

Yes. Checkpoint reviews occur after cell-level QC, after batch correction, and before final delivery when scoped at kickoff. You can review cluster annotations, SCR filtering thresholds, contrast definitions, and marker-protein tables within agreed scope. A dedicated scientific contact leads the project, coordinates milestone feedback, and documents decisions in the QC report and Methods draft.

What does post-delivery reviewer support include?

Clarification of methods, QC thresholds, SCR filtering, and minor figure or table revisions within agreed scope (typically ≤20% of project scope), consistent with Pepkio's standard post-delivery policy. It does not include open-ended reanalysis or new biological contrasts. Substantial new work such as additional contrasts, alternate imputation, or scRNA-seq integration is scoped as separate milestones.

Is co-authorship required?

No. Pepkio is fee-for-service unless co-authorship is agreed in advance. You retain ownership of results and code. Billing is separate from publication credit.

How should I design the isobaric carrier ratio for TMT-based single-cell proteomics?

Carrier amount should reflect project priorities. Isobaric carriers about 200-fold larger than single-cell samples enhance peptide identification with a limited adverse effect on quantification [11]. Larger carrier ratios can increase identification depth but can also compress single-cell dynamic range and worsen TMT channel impurity effects [12]. Pepkio documents carrier design assumptions and SCR thresholds in the QC report. Channel exclusion is applied when impurity correction is insufficient.

Should I use TMT carrier multiplexing, label-free DIA, or plexDIA for single-cell proteomics?

TMT with an isobaric carrier (SCoPE2-style) quantified about 1,000 proteins per cell in published benchmarks [1]. Label-free DIA and plexDIA can increase depth. nPOP reports about 3,000-3,700 proteins per cell with plexDIA [7], but those workflows need workflow-specific informatics [13]. Pepkio documents trade-offs at kickoff based on your export format and biological question.

Can Pepkio run custom or non-standard single-cell proteomics analyses?

Yes, when a feasibility review confirms inputs, outputs, and timeline. Custom matrices, client QC thresholds, scRNA-seq integration, trajectory inference, phosphoproteomics, or non-standard deliverables are scoped at kickoff. Raw .raw reprocessing can be scoped via the DDA/DIA proteomics service when needed.

Related services

  • DDA/DIA proteomics — Bulk cohort quantification or spectral library building when single-cell depth is not required.
  • Olink proximity extension — Targeted plasma or CSF validation of MS-discovered protein signatures.
  • Single-cell RNA-seq — Matched transcriptome profiling when sample IDs align for proteogenomic correlation.
  • Multi-omics integration — Cross-layer modeling when proteomics, transcriptomics, and metabolomics datasets share sample identifiers.
  • Custom analysis — Non-standard SCP extensions, phosphoproteomics, or client-specified toolchains beyond the standard pipeline.
References
  1. Specht H, Emmott E, Petelski AA, et al. Single-cell proteomic and transcriptomic analysis of macrophage heterogeneity using SCoPE2. Genome Biology. 2021;22(1):50. https://doi.org/10.1186/s13059-021-02267-5 (PMID: 33504367)
  2. Deutsch EW, Orsburn BC, Aebersold R, et al. Initial recommendations for performing, benchmarking and reporting single-cell proteomics experiments. Nature Methods. 2023;20(3):375–386. https://doi.org/10.1038/s41592-023-01785-3 (PMID: 36864200)
  3. Rosenberger FA, Thielert M, Strauss MT, et al. Spatial single-cell mass spectrometry defines zonation of the hepatocyte proteome. Nature Methods. 2023;20(10):1530–1536. https://doi.org/10.1038/s41592-023-02007-6 (PMID: 37783884)
  4. Vanderaa C, Gatto L. Revisiting the thorny issue of missing values in single-cell proteomics. Journal of Proteome Research. 2023;22(9):2775–2784. https://doi.org/10.1021/acs.jproteome.3c00227 (PMID: 37530557)
  5. Demichev V, Messner CB, Vernardoeil S, Lilley KS, Ralser M. DIA-NN: neural networks and interference correction enable deep proteome coverage in high throughput. Nature Methods. 2020;17(1):41–44. https://doi.org/10.1038/s41592-020-00998-0 (PMID: 31768060)
  6. Cox J, Mann M. MaxQuant enables high peptide identification rates, individualized p.p.b.-range mass accuracies and proteome-wide protein quantification. Nature Biotechnology. 2008;26(12):1367–1372. https://doi.org/10.1038/nbt.1511 (PMID: 19029910)
  7. Leduc A, Derks J, Slavov N. Massively parallel sample preparation for multiplexed single-cell proteomics using nPOP. Nature Protocols. 2024;19(12):3750–3776. https://doi.org/10.1038/s41596-024-01033-8 (PMID: 39117766)
  8. Lazar C, Gatto L, Ferro M, Bruley C, Burger T. Accounting for the multiple natures of missing values in label-free quantitative proteomics data sets to compare imputation strategies. Journal of Proteome Research. 2016;15(4):1116–1125. https://doi.org/10.1021/acs.jproteome.5b00981 (PMID: 26906401)
  9. Korkmaz S, Cox J, Grosse I, et al. Accurate and robust Bayesian inference of proteome-wide differential expression. Nature Methods. 2020;17(12):1215–1221. https://doi.org/10.1038/s41592-020-00949-9
  10. Ritchie ME, Phipson B, Wu D, et al. limma powers differential expression analyses for RNA-sequencing and microarray studies. Nucleic Acids Research. 2015;43(7):e47. https://doi.org/10.1093/nar/gkv007 (PMID: 25605792)
  11. Leduc A, Huffman RG, Specht H, et al. Optimizing accuracy and depth of protein quantification in experiments using isobaric carriers. Journal of Proteome Research. 2020;19(11):4356–4365. https://doi.org/10.1021/acs.jproteome.0c00675 (PMID: 33190502)
  12. Cheung TK, Koller A, Slavov N. A deeper look at carrier proteome effects for single-cell proteomics. Communications Biology. 2022;5:1170. https://doi.org/10.1038/s42003-022-03095-4 (PMID: 35194133)
  13. Wang J, Huang Y, Lu F, et al. Benchmarking informatics workflows for data-independent acquisition single-cell proteomics. Nature Communications. 2025;16:10276. https://doi.org/10.1038/s41467-025-65174-4 (PMID: 41271703)

Let's Talk About Your Science

Tell us:

  • • Your biological question
  • • Data type and size
  • • Timeline constraints

We'll tell you:

  • • What's feasible
  • • How long it will take
  • • Exactly what it will cost
Discuss Your Project

Contact us to start with a free consultation. Need everyday bench calculators? Try our free lab tools.