CASCADE

CASCADE , or Cancer Aggressiveness via Single-Cell Analysis During Evolution, is a bioinformatics tool that quantitatively analyzes liver tumors from single-cell or bulk transcriptomic data. It distills the tumor landscape down into 3 metrics:

• Lineage score: representation of the malignant cell community. Ranges from 0 to 1, where a 0 indicates a hepatocyte- or myeloid cell-like tumor lineage and a 1 indicates a cholangiocyte- or mesenchymal-like tumor lineage.
• Ecological score: representation of the tumor microenvironment. Ranges from 0 to 1, where a 0 indicates a restrictive tumor microenvironment and a 1 indicates a lenient one.
• Tumor score: the product of the lineage and ecological score.

The intersection of the lineage and ecological scores separates samples into 4 quadrants based on whether they are greater than or less than 0.5 along each metric. These quadrants are termed as follows:

• A1: lineage score < 0.5 and ecological score < 0.5
• A2: lineage score < 0.5 and ecological score >= 0.5
• B1: lineage score >= 0.5 and ecological score < 0.5
• B2: lineage score >= 0.5 and ecological score >= 0.5

The tumor score is an alternative to the lineage and ecological scores that distills this quadrant system down to a single number.

Citation

TBD

Install

Install CASCADE directly from GitHub:

library(devtools)
install_github("MaLab621/CASCADE")
library(CASCADE)

Usage

CASCADE works with either single-cell or bulk transcriptomic data. For single-cell data, input either a single Seurat object containing malignancy information in the metadata or separate Seurat objects for malignant and nonmalignant cells. Both forms of input also require metadata columns for sample identity and cell type. For bulk data, input a normalized matrix with genes on the rows and samples on the columns.

Running with a single Seurat object:

obj <- CreateSeuratObject(counts = gene_counts)
obj$malignancy_label <- c("malignant", "non-malignant", "non-malignant")  # your label vector would be much longer than this
obj$sample_name <- c("s1", "s1", "s2")                                    # your label vector would be much longer than this
obj$cell_type <- c("tumor cells", "B cells", "CD8-GZMH-effector T cells") # your label vector would be much longer than this

cascade_df <- sc_CASCADE(total_obj = obj, malignancy_col = "malignancy_label", 
                         ident_col = "sample_name", subtype_col = "cell_type")

Running with separate malignant and non-malignant objects:

malignant_obj <- CreateSeuratObject(counts = malignant_gene_counts)
malignant_obj$sample_name <- c("s1", "s2", "s3")                        # your label vector would be much longer than this

nonmalignant_obj <- CreateSeuratObject(counts = nonmalignant_gene_counts)
nonmalignant_obj$sample_name <- c("s1", "s4", "s5")                     # your label vector would be much longer than this
nonmalignant_obj$cell_type <- c("MAIT", "B cells", "CD4-KLRB1-T cells") # your label vector would be much longer than this

cascade_df <- sc_CASCADE(tumor_obj = malignant_obj, nontumor_obj = nonmalignant_obj,
                         ident_col = "sample_name", subtype_col = "cell_type", min_n_cells = 10)

Running with bulk data:

cascade_df <- bulk_CASCADE(bulk_expr = bulk_data)

Sample visualization plot:

ggplot(cascade_df, aes(x = lineage_score, y = eco_score)) +
  geom_point() +
  geom_hline(yintercept = 0.5, linetype = "dashed") +
  geom_vline(xintercept = 0.5, linetype = "dashed") +
  xlim(c(0,1)) +
  ylim(c(0,1)) + 
  labs(x = "Lineage score", y = "Ecological score") +
  theme_bw() +
  theme(aspect.ratio = 1)

Usage notes

CASCADE is specifically designed to analyze liver tumors. As such, it contains very specific requirements on how inputted data should look. For single-cell data, only the following 25 cell type labels will be considered in the calculation of the ecological score:

B cells
Plasma cells
CD4-CD69-memory T cells
CD4-FOXP3-regulatory T cells
CD4-IL7R-central memory T cells
CD4-KLRB1-T cells
CD8-CD69-memory T cells
CD8-GZMH-effector T cells
CD8-GZMK-effector memory T cells
MAIT
NK-CD160-tissue resident
NK-GNLY-circulatory
T cells-MKI67-proliferative
c0-LUM-inflammatory CAF
c1-MYH11-vascular CAF
c2-APOA1-hepatocyte like CAF
c0-S100A8-Monocyte
c1-CXCL10-M1 Macrophage
c2-CCL4L2-M2 Macrophage
c3-TPSB2-Mast cells
c0-VWF-endothelial
c1-ANGPT2-endothelial
c2-CRHBP-endothelial
c3-CCL5-endothelial
c4-RGS5-endothelial

Your data does not need to contain all 25 of these cell types. You can label cells as types not included in this list, but those cells will be excluded from all score calculations. Make sure that labels are an exact match in spelling and case.

Additionally, the lineage score bases its calculations on the expression of genes related to 4 specific tumor cell lineages: Cholangiocyte-like tumor cells, Hepatocyte-like tumor cells, Myeloid cell-like tumor cells, and Mesenchymal-like tumor cells. No action is required to accomodate this functionality, but do consider if these lineages are appropriate for your dataset.