Fig 1.
MouSSE (Mouse-Specific Single cell transcriptomics level cytokine activity prediction and Estimation) supports both positive and negative gene set scoring with the modified VAM method constructed using each of the 86 cytokine-specific gene sets.
Fig 2.
Proportion of cytokines with the highest Area Under the Receiver Operating Characteristic Curve (AUC-ROC) score calculated from cross-validated mouse lymph node target scRNA-seq data, encompassing 77,033 cells.
(Fig 2A) Comparison of MouSSE scores (mousse) with other signaling estimation methods, including normalized ligand (naive) and receptor scores (receptor), ligand-receptor product (product), NICHES application (niches), and PROGENy application (progeny). (Fig 2B) Evaluation of MouSSE scores relative to alternative gene set construction approaches, such as IREA-based gene sets scored using only positively weighted genes (irea.pos) and Seurat-based perturbation scores (seurat). (Fig 2C) Sensitivity analysis of MouSSE’s gene set construction and scoring methodology, comparing it with IREA-based gene sets incorporating both positive and negative weights and scored using MouSSE’s weighting strategy (irea.mousse) and modified MouSSE approach using only positively weighted genes (mousse.pos) and using only negatively weighted genes (mousse.neg). (Fig 2D) Overall comparison of MouSSE scores across all ten methods.
Fig 3.
Average Area Under the Receiver Operating Characteristic Curve (AUC-ROC) for all 86 cytokines quantified using the MouSSE and comparative methods as computed on the 77,033 cell mouse lymph node target scRNA-seq data.
Cytokine markers with an asterisk have the highest AUC-ROC score when estimated using the MouSSE method.
Fig 4.
Proportion of cytokines with the highest AUC-ROC, PR-AUC, balanced accuracy, detection rate, sensitivity, detection prevalence, F1 score, NPV, precision, specificity and prevalence across each of the 10 methods as computed from cross-validated mouse lymph node target scRNA-seq data, encompassing 77,033 cells.
Fig 5.
Minimum absolute avg_log2FC threshold for each of the 86 cytokines for gene set size ranging from 20, 60, 100 and 200 genes as computed from cross-validated mouse lymph node target scRNA-seq data.
Sum of thresholds over all gene set size is indicated for each cytokine.
Fig 6.
Differentially expressed cytokine markers on scRNA-seq data consisting of 63,734 samples by COVID19 severity condition (i.e., control, mild and severe) for the MouSSE method.
Fig 7.
Heatmap of Pearson correlation between cytokines differentially expressed by COVID19 severity as estimated using the MouSSE method (see Fig 6) and the Hallmark-specific pathway scores as computed on the COVID19 scRNA-seq data consisting of 63,734 samples.
Fig 8.
Differentially expressed cytokine markers on mouse lymph node ST data consisting of 1092 spots by infection status (i.e., MS-1, MS-2 and PBS).
(Fig 8A) Heatmap of differentially expressed cytokines for the MouSSE method (mousse). (Fig 8B) Heatmap of differentially expressed cytokines for the modified MouSSE method scored on just the positive weighted genes (mousse.pos).
Fig 9.
Differentially expressed cytokine markers on mouse lymph node ST data consisting of 1092 spots by infection status (i.e., MS-1, MS-2 and PBS).
(Fig 9A) Heatmap of differentially expressed cytokines for the IREA-based gene sets scored using just positive weighted genes (irea.pos). (Fig 9B) Heatmap of differentially expressed cytokines for the IREA-based positive and negative weighted gene sets scored using MouSSE’s weighting strategy (irea.mousse).