Cross-platform cytometry benchmark data

This repository contains the data presented in our original article, "Superior Precision of Clinical Predictions after CD3-relativisation to Align Flow Cytometry Data," including raw and processed FCS files from 482 samples. This dataset captures information about T cell distributions in human healthy donors through standardised flow cytometry measurements made in four internationally collaborating laboratories over 17 months using 6 different cytometers. A subset of 329 samples split for parallel measurements on at least 2 instruments. This repository also reports donor-level clinical and demographic information, as well as QC files. If you want to start analysing the data, we propose you download CORE.06-1_rel.asinhCD3 which are CD3-relativised and properly arcsinh transformed data of the CORE studies R1, R2 and SAN. In addition, download patientdata.zip, where all anonymized patient information necessary is reported. See technical note below if you want to use CORE.02-COMPENSATED.7z! Cohort overview CORE: Our core dataset comprises 459 samples from 358 unique donors, which are organized into three main cohorts (R1, R2 and SAN). R1 includes 254 samples from 153 unique donors, which were analysed in Regensburg. Clinical and demographic variables are recorded. Within R1, 101 repeated samples were collected from 60 donors to test the biological stability of T cell subset distributions over time. The first samples taken from each R1 donor were randomised into training (50), validation (50) and test (53) sets. R2 is a prospective dataset collected 4 months after R1 that includes 52 samples from unique donors, who are not represented in R1. R1 and R2 samples were split after staining and measured in parallel with a Navios™ (Navios) and CytoFLEX LX™ (LX) cytometer. SAN comprises 153 samples from unique donors that were measured with a DxFlex™ (DxFLEX) cytometer in Santander. As with R1, SAN samples were randomised into training (50), validation (50) and test (53) sets. EXTENDED: Our extended dataset incorporates two cohorts, FORT and BAD. FORT comprises 14 samples from unique donors that were split into equal parts, then measured in parallel using 3 cytometers – namely, a Navios™ and CytoFLEX LX™ from Beckman Coulter, and an LSRFortessa™ (LSR) from Becton Dickenson. The BAD cohort incorporates samples from 9 unique donors that were split after staining and measured in parallel using an LSRFortessa™ and a Cytek Aurora™ (CA) spectral cytometer. Processing overview For each cohort, we report (a subset of) the data in the following processing stages, denoted as e.g. CORE.01_RAW. The subsets -1 and -2 include different staining panels after gating them to "useful" cells, where -1 is the thing you probably want for analysis as these are the samples stained with all colors together gated to T cells based on CD3+. 01_RAW: Raw, untransformed data for full and single stained samples. Only marker renaming was performed to harmonize measurements from all cytometers. 02_COMPENSATED: Compensated FCS files, using by-sample manually curated spillover matrices. Only BAD cohort used untouched device-compensations. 03-1_gatedCD3: Singlets/CD45+ Leukocytes/CD3+ T cell gated samples. Only `...12-panel.fcs` are included. 03-2_gatedLympho: Lymphocyte gated samples based on forward and side scatter. All single stained and unstained samples (...01-CD3-FITC.fcs, until ...11-none.fcs) 04-1_asinhCD3 and 04-2_asinhLympho: Data from 03-1_gatedCD3 or 03-2_gatedLympho after arcsinh transformation, with different manually optimized arcsinh-cofactors per cytometer. 05-1_relativizedCD3 and 05-2_relativizedLympho: Data from 03-1_gatedCD3 or 03-2_gatedLympho after applying sample-wise relativisation. 06-1_rel.asinhCD3 and 06-2_rel.asinhLympho: Data from 05-1_relativizedCD3 and 05-2_relativizedLympho after applying one global arcsinh transformation Patient information Can be found in patientdata.zip. Content is the patient information of BAD_pheno_processed.csv: BAD cohort FORT_pheno_processed.csv: FORT cohort pheno_full_processed.csv: Complete CORE cohort R1_pheno_first_processed.csv: R1 samples from a donor's first presentation R1_pheno_processed.csv: All R1 samples+patients R2_pheno_processed.csv: R2 cohort SAN_pheno_processed.csv: SAN cohort Gatings For CORE and FORT cohort samples the gating strategies are manually curated for each sample on compensated, untransformed files - for full, single and unstained samples. They are supplied as flowWorkspace GatingSets and we have applied them sample by sample. The gating strategy is always the same, just the gate positions have been curated. For BAD samples, we have one gating strategy for all samples per cytometer.