Table 1.
Patient and control characteristics.
The demographics and symptom levels of the patients used in this study.
Fig 1.
The characteristics of the patients.
A heatmap of the clinical scores for the 133 patients included in this study. The values have been scaled between zero (absent) and one (worst). ESSDAI = EULAR Sjögren’s Syndrome Disease Activity Index, SSDDI = Sjögren’s Syndrome Disease Damage Index, ESSPRI = EULAR Sjögren’s Syndrome Patient Reported Index, HAD = Hospital Anxiety and Depression, PROFAD = Profile of Fatigue and Discomfort, VAS = Visual Analogue Scale.
Fig 2.
Differential gene expression analysis.
(A) Volcano plot of high fatigue against low fatigue. No significant differentially expressed genes (DEGs) were detected. (B) Volcano plot of patients against healthy controls. Red points indicate DEGs with a fold change >1.2 and p-value <0.05. (C) The mean expression values for each gene for the high and low fatigue groups. (D) Plot of the first two principal components of the expression dataset coloured by high and low fatigue groups.
Fig 3.
Correction for other clinical factors.
Volcano plots for the Fatigue VAS fatigue groups corrected for clinical factors: (A) Age at UKPSSR cohort recruitment. (B) Disease activity measured using the EULAR Sjögren’s Syndrome Disease Activity Index. (C) Disease damage measured using the Sjögren’s Syndrome Disease Damage Index. (D) The EULAR Sjögren’s Syndrome Patient Reported Index dryness sub-domain. (E) The EULAR Sjögren’s Syndrome Patient Reported Index pain sub-domain. (F) Anxiety measured using the Hospital Anxiety and Depression scale. (G) Depression measured using the Hospital Anxiety and Depression scale. (H) Pain and depression (E & G). (I) Pain, depression, dryness and anxiety (D-G). (J) All seven factors (A-G). No significantly differentially expressed genes were identified following any correction.
Fig 4.
Interferon type I signature and fatigue.
(A) The IFN score ranges for the 133 patients. (B) The Fatigue VAS scores for the IFN-active and IFN-inactive groups. (C) The ESSDAI scores for the IFN-active and IFN-inactive groups.
Table 2.
Enriched pathways between the Fatigue VAS high fatigue and low fatigue groups.
Gene sets were considered to be enriched at an FDR cut-off of 25%. All the enriched gene sets were associated with high fatigue with the exception of incretin synthesis secretion and inactivation (*), which had a non-random distribution of enriched genes between the two fatigue groups.
Table 3.
Genes in the leading edge of the enriched actin-related BioCarta pathways.
Genes found in leading edge overlap are shown in bold.
Table 4.
Genes in the leading edge of the enriched Reactome G-protein signalling pathways.
Genes found in leading edge overlap are shown in bold.
Table 5.
Genes in the leading edge of the incretin-related Reactome pathway.
Genes associated with high fatigue are shown in bold.
Fig 5.
Support vector machine (SVM) classification of fatigue groups.
The receiver operator characteristic curves for the SVM output. Ten curves are shown on each plot. The area under the curve (AUC) is calculated as the mean over the ten curves. (A) All 181 enriched pathway genes as input. (B) The 55 leading edge genes as input.
Fig 6.
A workflow of the gene expression analysis.
The gene expression data were analysed to produce a list of fatigue-related features which were used as inputs for a support vector machine classifier of fatigue. 1. Differentially expressed genes were identified between fatigue groups. 2. Linear regression was used to analyse fatigue as a continuous variable. 3. The interferon type I signature was calculated for all the patients and compared to fatigue levels. 4. Gene set enrichment analysis was carried out using the high and low fatigue groups. 5. A support vector machine classifier was created using fatigue-related features as inputs and its performance assessed using receiver-operator characteristic (ROC) curves.