The dataset includes two tabular sets of concentration measurements of selected proteins in plasma samples from patients with soft tissue infections or sepsis. Patients with soft tissue infections are classified into three groups: Patients with necrotizing soft tissue infections (NSTI), suspected NSTI cases but where no necrotic tissue was found during surgical exploration (Non-NSTI) and cellulitis. The sepsis patient cohort has heterogeneous etiologies and location of infection. Additionally, we included a group of patients that had surgery related to non-infectious conditions as a healthy cohort of patients (Surgical control).

The measurements were performed by Luminex® multiplex immunoassay or ELISA. Each of the two sets were obtained from independent rounds of measurements and differ in the panel of analytes measured and samples included. The first set of measurements (Set01) covers 39 analytes measured in two sets of customized multiplex plates (32-plex and 5-plex), and two in independent ELISA assays. These analytes were measured in 251 NSTI samples, 20 Non-NSTI, 19 cellulitis and 20 surgical controls. The second set of measurements (Set02) consist of a subset of 10 analytes included in the first set and were measured in two multiplex plates (4-plex and 6-plex). The second round of measurements were carried out in 60 additional NSTI patients and 24 sepsis patients.

Lastly, imputation of censored data was carried out only in the first set of measurements and the resulting data is also available in the current data set (Set01_imputed).

The study is based on clinical data and plasma samples from patients with NSTI (surgically confirmed) enrolled in the multicentre INFECT study (NCT01790698, clinicaltrials.gov). Samples were collected in five hospitals in Scandinavia: Blekingesjukhuset (Karlskrona, Sweden), Haukeland University Hospital (Bergen, Norway), Karolinska University Hospital (Stockholm, Sweden), Righospitalet (Copenhagen, Denmark), Sahlgrenska University Hospital (Gothenburg, Sweden). The ‘non-NSTI’ patient samples were collected during the INFECT study and included patients with suspected NSTI who after surgical examination, were diagnosed with less severe soft tissue infections due to lack of necrotic tissue.

The ‘surgical controls’ included patients who underwent elective surgery at Rigshospitalet (University Hospital of Copenhagen, Denmark) for non-infectious conditions and that had no underlying diseases. Plasma samples from patients with sepsis at the emergency clinic at the Karolinska University Hospital Huddinge were included.

Plasma samples from patients were prepared from blood collected at admission in EDTA-containing tubes and immediately aliquoted and frozen at -80°C. Concentrations in plasma of the selected panels of analytes were determined using the bead based Luminex multiplex immunoassay. Samples were thawed on the day of measurement and assays were performed according to the manufacturer’s protocol. All plates were acquired on a Luminex MAGPIX instrument using xPonent 4.0 Software (Luminex). The measurements of the first set of analytes (Set01) were done in two individual ELISA plates for IL-23 and IL-33, and two customized multiplex plates of 5 and 32 analytes (R&D systems) including  CCL2/MCP-1, CCL4/MIP-1β, CCL5/RANTES, CXCL-8/IL-8, CXCL10/IP-10, IL-1α, IL-1β, IL-2, IL-4, IL-6, IL-10, IL-12p70, IL-13, IL-17A, IL-18, IL-22, IL-36β/IL-1F8, E-Selectin, ICAM-1, VCAM-1, MMP-1, MMP-8, MMP-9, C5/C5a, Collagen-IVα1, Fas-Ligand, Galectin-3, G-CSF, I-alpha-1/COL1A1, MPO, Pentraxin-3, Resistin, S100A8, S100A9, Thrombomodulin, and TNFα.

Set02 consist of two panels measured in two customized multiplex plates from R&D Systems (G-CSF, IL-6, S100A8 and Thrombomodulin) and Thermofisher (MMP-9, CXCL10, IL-2, IL-10, IL-22, and Fas-Ligand).

Imputation of censored data was only carried out in Set01. Censored values from cytokines with only one missing value were substituted directly to half the minimal value (left-censored) or the maximal plus 20% (right-censored). For all other cytokines, the imputation was performed using the method by Wei et al. (PLoS Comput. Biol. 14, 2018, doi:10.1371/journal.pcbi.1005973.). The imputation for cytokines with double censored data was carried out in two sequential imputation steps. First, the data was made left-censored by setting the censored values above the range to the maximum observed value and then performing imputation of the left-censored data. Then, the right-censored values were set back to be censored and imputed.