Psoriatic arthritis (PsA) clinical lipidomics dataset with hidden laboratory workflow artifacts

This dataset provides plasma lipidomic profiles from a cross-sectional cohort of patients with psoriatic arthritis (PsA) and healthy controls matched by age, sex, and BMI. Plasma samples were collected at University Hospital Frankfurt and the German Red Cross Blood Transfusion Service in Frankfurt, Germany, and analyzed on high-resolution mass spectrometry platforms (Thermo Fisher Q Exactive and Sciex QTRAP 6500+) using both targeted and untargeted lipid assays.

The dataset is organized into four comma-separated values (CSV) files containing Box-Cox-transformed and imputed lipidomics values, corresponding back-transformed measurements on the original scale, detailed clinical and analytical metadata including PsA status, ESOM-based classes, sex, and assay-specific batch identifiers, and variable-level descriptions for all 292 lipids.

Across all files, 292 plasma lipid species are included, named according to LIPID MAPS classification and standardized lipid nomenclature. The variables span major lipid classes such as carnitines, ceramides, glycerophospholipids, sphingolipids, glycerolipids, fatty acids, sterols and esters, and endocannabinoids.

Because the dataset contains an embedded batch structure, it is not intended for deriving new biological conclusions. Instead, it serves as a resource for methodological research on data quality control, batch effect identification, evaluation of preprocessing pipelines, and assessment of how sampling and processing parameters influence analytical robustness.

The dataset consists of four CSV files:

"PsA_lipids_BC.csv"
Contains the primary lipidomics data matrix with 107 subjects and Box-Cox-transformed, outlier-cleaned, and imputed plasma lipid measurements for 292 lipids.

"PsA_lipids_orig_values.csv"
Contains the same subjects and variables as above, but values are back-transformed to the original measurement scale following outlier removal and imputation, preserving identical identifiers and structure.

"PsA_classes.csv"
Provides 107 rows and 12 columns of metadata linking each subject to clinical (PsA vs. control), ESOM-based, and gender classifications, as well as six assay-specific batch identifiers. Columns 11 and 12 include the sampling date and weekday for control samples, enabling investigation of temporal or workflow-related effects.

"readme.csv"
Contains 292 rows (matching the number of lipid variables) and 7 columns describing each lipid at the individual variable level: "variable_name" (lipid identifier), "unit" ("arbitrary" for screening = peak area relative to class-specific internal standard; "ng/mL" or "pg/mL" for targeted), "class_name" (e.g., "Fatty acids", "Lysophosphatidylcholine"), "class_code" (e.g., "FA", "LPC"), "analytical_method_category" ("Lipid screening" or "Lipid targeted"), "LLOQ", and "ULOQ" (quantification limits; NA for screening lipids).