Fig. 6: Validation of SPIDER on identifying TF-supported SVIs on datasets at bulk resolution.

A–D. Validation of SPIDER on identifying TF-supported SVIs on multiple breast cancer samples at the bulk resolution. A The boxplot of five SVI evaluation metrics and one TF correlation metric (n=490, 487, 2117, 944, 563, and 647 LRIs for samples A1 to G2, respectively). B Marker SVIs for the main clusters involved in sample G2 with correlation coefficients higher than 0.3. C Marker SVIs for invasive cancer and cancer in situ that are consistent across samples, shown on samples A1 and G2. D The dot plot lists pathways enriched by genes implicated in constant cancer marker SVIs. E Pseudotime results based on SVIs (left) and gene expression (right) shown on samples A1 and G2. F The barplot of correlations between cancer/TME labels and pseudotime from SVI or gene expression across all samples. G Boxplots showing pseudotime distributions with respect to cancer/TME labels (n=307, 140, 20 spots for TME, Invasive Cancer, and Cancer In Situ, respectively). H–L Validation of SPIDER on identifying TF-supported SVIs on multiple DLPFC samples at the bulk resolution. H The boxplot of five SVI evaluation metrics and one TF correlation metric (n=1584, 1416, and 1454 LRIs for sample 151673, 151510, and 151672, respectively). I Correlation heatmap of white matter and layer 3 regions on three samples with top three white matter marker SVIs (left) and layer 3 marker SVIs (right). The white color indicates the corresponding SVI is missing in the sample. J The dot plot lists pathways enriched by genes implicated in constant region marker SVIs. K Trajectories inferred with gene expression (top) and SVIs (bottom) on sample 151673. L The barplot showing AUROC scores of trajectories inferred from gene expression and SVIs on three samples. Corr: Pearson correlation coefficient; WM: white matter. The boxplots display the median (center line), the 2 and 75th percentiles (box bounds), whiskers extending to the most extreme data points within 1.5 × the interquartile range, minima and maxima as the lowest and highest points within the whiskers, and outliers as individual points beyond the whiskers. The statistical significance of box plots is calculated using one-sided Mann-Whitney-Wilcoxon test with Benjamini-Hochberg correction, with the exact adjusted p-values listed in Supplementary Table 8 and the following significance annotations: ****: adjusted p-value ≤ 1.00e-04; ***: 1.00e-04  < adjusted p-value ≤ 1.00e-03; **: 1.00e-03  < adjusted p-value ≤ 1.00e-02; *: 1.00e-02  < adjusted p-value ≤ 5.00e-02. Source data are provided as a Source Data file.