Fig. 3: Performance assessment of scMORE using ten hematological traits.
a, Benchmarking analysis of three GRN inference methods (GLM, XGBoost and GLMNET) for scMORE across 10 hematological traits in the ground-truth dataset. Both cosine similarity and averaged gene expression methods were evaluated for each GRN method. A two-sided paired Student’s t-test was used to assess statistical significance (n = 10 blood traits). Each dot in boxplot indicates a blood trait. Boxplot inside the violin plot shows the median (center line), IQR (box) and 1.5 × IQR bounds (whiskers). Minima and maxima are represented by the whiskers. b. Consistent identification of seven eRegulons relevant to lymphocyte count and percent across the three GRN inference methods. **Denotes PCTS < 0.05, PGRS < 0.05 and PTRS < 0.01. *Denotes PCTS < 0.05, PGRS < 0.05 and PTRS < 0.05. #Denotes PCTS < 0.05, PGRS < 0.1 and PTRS < 0.05. Significance was assessed by an MC permutation test (one-sided, upper tail; n = 1,000 iterations; no multiple correction). c, Two-sided Pearson correlation analysis of TRSs between scMORE-based GLM and GLMNET, as well as XGBoost. d, Benchmarking analysis of scMORE’s cosine specificity performance compared to averaged gene expression and MAGMA_Cell-typing-based gene specificity across ten hematological traits using five immune single-cell multiomic datasets. These datasets included varying cell counts (1 K, n = 1,222 cells; 3 K, n = 2,722 cells; 5 K, n = 5,174 cells; 7 K, n = 7,304 cells; 9 K, n = 8,900 cells), spanning seven cell types (monocytes, mDCs, pDCs, B cells, NK cells, CD8+ T cells and CD4+ T cells). The E-statistics for each trait was calculated using the getEnergyScore() function in scMORE. Boxplot shows the median (center line), IQR (box) and 1.5 × IQR bounds (whiskers). Minima and maxima are represented by the whiskers. Each dot in boxplot indicates a blood trait (n = 10 blood traits).
