Defining Quality Control Standards for Single-Cell Proteomics by Inter-Laboratory Benchmarking

Résumé fourni par la source

This deposit contains data associated with a multi-laboratory single-cell proteomics benchmarking study. Nine laboratories contributed data acquired on Bruker timsTOF and Thermo Fisher Astral instruments. Samples consisted of single HEK293T and K562 cells, as well as bulk carrier-free mixtures of the two cell types. ABSTRACT: Single-cell proteomics can quantify thousands of proteins from individual mammalian cells, yet the absence of community-wide quality control limits biological interpretability. Here, the HUPO Single Cell Initiative presents the first inter-laboratory single-cell proteomics benchmarking study across seven laboratories using standardized 384-well plates acquired on Orbitrap Astral and timsTOF Ultra2 instruments. Centralized analysis across six DIA software tools revealed that software choice impacts identification depth and quantitative accuracy more than instrument vendor. Multi-layered quality control enabled the detection of cell-leakage during sorting, LC misconfiguration, column degradation and site-specific pipetting failures. Quantitative correlations were driven by sample-type and vendor while Astral-ABIRD single-cells correlated more closely with timsTOF than Astral-FAIMS instruments. Sequential correction for plate identity and well position recovered clean cell-type separation for confident downstream differential expression analysis. This study provides a data-driven quality control framework spanning plate design to batch correction for reproducible single-cell proteomics across laboratories and platforms. --- ## Deposit contents The data are organized into the following archives: | Archive | Size | Contents | |---|---|---| | `cellenONE.zip` | ~714 MB | cellenONE cell dispensing run files (14 plates) with images| | `metadata.zip` | ~2 MB | SDRF metadata files per laboratory | | `search_results_Human.zip` | ~21 GB | DIA-NN search results, single-cell Human samples (9 labs) | | `search_results_Mixture.zip` | ~22 GB | DIA-NN + Spectronaut search results, Mixture samples (9 labs) | | `Data_AnnData.zip` | ~29 GB | AnnData objects (precursor and protein level, DIA-NN and Spectronaut) | | `Data_scplainer.zip` | ~5.7 GB | scplainer intermediate and result objects | | `Data_MSqRob.zip` | ~190 MB | MSqRob differential abundance results | | `Data_Proteobench.zip` | ~329 MB | Search engine comparison intermediates and datapoints | > **Note:** Spectronaut search results for the Human single-cell samples are hosted on PRIDE > (see `search_results/Human/Spectronaut/README.md`). --- ## Folder structure ``` . ├── cellenONE/ │ ├── Plate_01_Backup/ │ │ ├── HEK_plate1_*.Run │ │ ├── K562_plate1_*.Run │ │ └── NEG_plate1_*.Run │ └── Plate_02_BOGDAN_/ ... Plate_14_ERWIN/ (same structure, 14 plates total) │ ├── metadata/ │ └── SDRF/ │ ├── Broad_timsTOF_1/SDRF.xlsx │ ├── BYU_timsTOF_1/SDRF.xlsx │ ├── DTU_Astral_1/SDRF.xlsx │ ├── IMP_Astral_1/SDRF.xlsx │ ├── IMP_timsTOF_1/SDRF.xlsx │ ├── KI_Astral_1/SDRF.xlsx │ ├── KI_Astral_2/SDRF.xlsx │ ├── LSMBO_Astral_1/SDRF.xlsx │ └── Wyss_Astral_1/SDRF.xlsx │ ├── search_results/ │ ├── Human/ │ │ ├── DIANN/ │ │ │ └── /DIANN_2.6/ │ │ │ ├── report.parquet │ │ │ ├── report-first-pass.parquet │ │ │ ├── empirical_speclib.parquet │ │ │ ├── empirical_speclib.parquet.skyline.speclib │ │ │ ├── report.stats.tsv │ │ │ ├── report.log.txt │ │ │ └── quant/ │ │ └── Spectronaut/ (hosted on PRIDE — not included here) │ └── Mixture/ │ ├── DIANN/ │ │ └── /DIANN_2.6/ (same structure as Human/DIANN) │ └── Spectronaut/ │ └── / │ ├── Report_BGS Factory Report (Normal).tsv │ ├── RunIdentifications.tsv │ ├── ConditionSetup.tsv │ ├── Peptide CVs_CVsBelowX.tsv │ ├── Precursor CVs_CVsBelowX.tsv │ ├── Protein CVs_CVsBelowX.tsv │ ├── Protein Group CVs_CVsBelowX.tsv │ ├── AnalysisLog.txt │ └── RunSummaries/ │ ├── Data/ │ ├── metadata/ │ │ └── metadata.parquet (sample-level metadata table) │ ├── AnnData/ │ │ ├── DIANN/ │ │ │ ├── adata_precursor.h5ad │ │ │ ├── adata_precursor_qc.h5ad │ │ │ ├── adata_protein.h5ad │ │ │ └── adata_protein_qc.h5ad │ │ └── Spectronaut/ │ │ ├── adata_precursor.h5ad │ │ ├── adata_precursor_qc.h5ad │ │ ├── adata_protein.h5ad │ │ ├── adata_protein_qc.h5ad │ │ └── scplainer_results.h5ad │ ├── scplainer/ │ │ ├── sce.rds (SingleCellExperiment, raw) │ │ ├── sce_norm.rds (after normalization) │ │ ├── sce_batch.rds (after batch correction) │ │ ├── sce_plate.rds (plate-level object) │ │ ├── sce_*_GMF.rds (GMF-corrected variants) │ │ ├── sce_resid*.rds (residual objects) │ │ ├── vaRes_*.rds (variance analysis results) │ │ ├── de_results.rds (differential expression results) │ │ └── scplainer_results.h5ad (final results in AnnData format) │ ├── MSqRob/ │ │ └── Spectronaut/ │ │ ├── QFeatures/ │ │ │ └── qf_checkpoint.rds (QFeatures object checkpoint) │ │ └── DEA_results/ │ │ ├── csv/ (differential abundance results, CSV) │ │ ├── rds/ (differential abundance results, RDS) │ │ └── per_plate/ │ │ ├── dea_Astral_*.csv │ │ ├── dea_timsTOF_*.csv │ │ ├── combined.csv │ │ └── results_per_plate.rds │ └── ProteoBench/ │ ├── datapoints.pkl (ProteoBench result datapoints) │ └── intermediates.pkl (ProteoBench intermediate objects) │ └── README.md ``` --- ## Laboratory identifiers | ID | Institution | Instrument | |---|---|---| | Broad_timsTOF_1 | Broad Institute | Bruker timsTOF | | BYU_timsTOF_1 | Brigham Young University | Bruker timsTOF | | DTU_Astral_1 | Technical University of Denmark | Thermo Fisher Astral | | IMP_Astral_1 | IMP Vienna | Thermo Fisher Astral | | IMP_timsTOF_1 | IMP Vienna | Bruker timsTOF | | KI_Astral_1 | Karolinska Institute | Thermo Fisher Astral | | KI_Astral_2 | Karolinska Institute | Thermo Fisher Astral | | LSMBO_Astral_1 | LSMBO Strasbourg | Thermo Fisher Astral | | Wyss_Astral_1 | Wyss Institute | Thermo Fisher Astral | --- ## Software versions - **DIA-NN**: 2.6 - **Spectronaut**: 20 - **scplainer**: 1.20.0 - **MSqRob**: 1.18.0 --- ## File format notes - `.parquet` — columnar data format readable with pandas (`pd.read_parquet`) or R (`arrow::read_parquet`) - `.h5ad` — AnnData format readable with `anndata` (Python) or `zellkonverter` (R/Bioconductor) - `.rds` — R serialized objects, readable with `readRDS()` in R - `.pkl` — Python pickle objects, readable with `pickle.load()` in Python - `.Run` — cellenONE proprietary run folder (cellenONE software required) - `.sne` — Spectronaut project/library file (Spectronaut software required) - `SDRF.xlsx` — Sample and Data Relationship Format, HUPO-PSI community standard

Ce résumé expose les affirmations des auteurs. BNTIC ne l’interprète pas comme une validation indépendante des résultats.