Figure 6: Surface proteins associated with GP activity
Ziang Zhang
Last updated: 2026-07-28
Checks: 7 0
Knit directory:
immgenT-GP-analysis/analysis/
This reproducible R Markdown analysis was created with workflowr (version 1.7.2). The Checks tab describes the reproducibility checks that were applied when the results were created. The Past versions tab lists the development history.
Great! Since the R Markdown file has been committed to the Git repository, you know the exact version of the code that produced these results.
Great job! The global environment was empty. Objects defined in the global environment can affect the analysis in your R Markdown file in unknown ways. For reproduciblity it’s best to always run the code in an empty environment.
The command set.seed(1) was run prior to running the
code in the R Markdown file. Setting a seed ensures that any results
that rely on randomness, e.g. subsampling or permutations, are
reproducible.
Great job! Recording the operating system, R version, and package versions is critical for reproducibility.
Nice! There were no cached chunks for this analysis, so you can be confident that you successfully produced the results during this run.
Great job! Using relative paths to the files within your workflowr project makes it easier to run your code on other machines.
Great! You are using Git for version control. Tracking code development and connecting the code version to the results is critical for reproducibility.
The results in this page were generated with repository version 0f5b5da. See the Past versions tab to see a history of the changes made to the R Markdown and HTML files.
Note that you need to be careful to ensure that all relevant files for
the analysis have been committed to Git prior to generating the results
(you can use wflow_publish or
wflow_git_commit). workflowr only checks the R Markdown
file, but you know if there are other scripts or data files that it
depends on. Below is the status of the Git repository when the results
were generated:
Ignored files:
Ignored: .DS_Store
Ignored: .claude/
Ignored: analysis/.DS_Store
Ignored: analysis/.Rhistory
Ignored: analysis/assets/.DS_Store
Ignored: captions/
Ignored: code/.DS_Store
Ignored: code/other/topic_flashier_20250212.R
Ignored: code/other/topic_wrapper_20250215_alldata_backfit.sh
Ignored: data
Ignored: experiments/
Ignored: figures/.DS_Store
Ignored: figures/Previous/.DS_Store
Ignored: figures/Previous/bits/.DS_Store
Ignored: figures/Previous/bits/Figure 1/.DS_Store
Ignored: figures/Previous/bits/Figure 2/.DS_Store
Ignored: figures/Previous/bits/Figure 3/.DS_Store
Ignored: figures/Previous/bits/Figure 4/.DS_Store
Ignored: figures/Previous/bits/Figure 6/.DS_Store
Ignored: figures/Previous/bits/Figure 7/.DS_Store
Ignored: figures/Previous/bits/Figure S1/.DS_Store
Ignored: figures/Previous/bits/Figure S2/.DS_Store
Ignored: figures/Previous/bits/Figure S3/.DS_Store
Ignored: figures/Previous/bits/Figure S6/.DS_Store
Ignored: figures/Previous/bits/Figure S7/.DS_Store
Ignored: figures/final-selected/.DS_Store
Ignored: figures/final-selected/Figure 1/.DS_Store
Ignored: figures/final-selected/Figure 2/.DS_Store
Ignored: figures/final-selected/Figure S4/.DS_Store
Ignored: log/
Ignored: output/Figure2/
Ignored: output/FigureS5/S5_cell_metadata.csv.gz
Ignored: plan/
Ignored: reorder/
Ignored: tables/
Ignored: tmp/
Note that any generated files, e.g. HTML, png, CSS, etc., are not included in this status report because it is ok for generated content to have uncommitted changes.
These are the previous versions of the repository in which changes were
made to the R Markdown (analysis/Figure6.Rmd) and HTML
(docs/Figure6.html) files. If you’ve configured a remote
Git repository (see ?wflow_git_remote), click on the
hyperlinks in the table below to view the files as they were in that
past version.
| File | Version | Author | Date | Message |
|---|---|---|---|---|
| Rmd | 0f5b5da | Ziang Zhang | 2026-07-28 | Align all figure captions with captions_20260728_final.docx |
| html | 58eab27 | Ziang Zhang | 2026-07-28 | Build site: Fig. 6i-k caption on the curated CD69 GP subset |
| Rmd | 589a0ed | Ziang Zhang | 2026-07-28 | Fig. 6i-k: say the 10 CD69 GPs are curated, not the top 10 |
| html | 3fc3789 | Ziang Zhang | 2026-07-27 | Republish all 24 pages |
| html | 1390a03 | Ziang Zhang | 2026-07-27 | Republish all 24 pages |
| html | adaef21 | Ziang Zhang | 2026-07-27 | Build site: panel fixes and PDF-derived assets |
| html | 5b19858 | Ziang Zhang | 2026-07-27 | Build site. |
| Rmd | ffe285c | Ziang Zhang | 2026-07-27 | Reorganize figures/ and untrack local-only exploration notes |
| Rmd | 91ee059 | Ziang Zhang | 2026-07-26 | Select each panel’s code block by name, not by line number |
| html | 91ee059 | Ziang Zhang | 2026-07-26 | Select each panel’s code block by name, not by line number |
| Rmd | 06552b8 | Ziang Zhang | 2026-07-14 | Update Figure 6b heatmap layout |
| html | 06552b8 | Ziang Zhang | 2026-07-14 | Update Figure 6b heatmap layout |
| html | 92021bf | Ziang Zhang | 2026-07-02 | Build site. |
| Rmd | db7a5cd | Ziang Zhang | 2026-07-02 | Recover 3 data-provenance gaps into pipeline scripts |
| html | 827c89b | Ziang Zhang | 2026-07-02 | Build site. |
| Rmd | 8ac7f9f | Ziang Zhang | 2026-07-02 | Add data provenance notes to each script; remove conversational |
| Rmd | f9db962 | Ziang Zhang | 2026-07-02 | Simplify layout: drop old code/script folders, rename |
| html | c6e5086 | Ziang Zhang | 2026-07-02 | Build site. |
| html | 5a79883 | Ziang Zhang | 2026-07-02 | Build site. |
| Rmd | 2b0e445 | Ziang Zhang | 2026-07-02 | Fix GitHub source links to point at the new |
| html | cf1d0ac | Ziang Zhang | 2026-07-02 | Build site. |
| Rmd | 06b2461 | Ziang Zhang | 2026-07-02 | Initial commit: immgenT-GP-analysis |
| html | 06b2461 | Ziang Zhang | 2026-07-02 | Initial commit: immgenT-GP-analysis |
All panels except (a) are produced by script/Figure6.R,
which shares its CITE-seq setup with Figure
S6 via code/R/citeseq_shared_setup.R. The code below is
shown for reference (not re-executed on this page, since this script
takes about a minute to run); the images are its pre-rendered
output.
Setup
library(ggplot2)
library(ggrepel)
library(dplyr)
library(patchwork)
library(pheatmap)
library(tidyr)
library(Matrix) # protein matrices are dgCMatrix; must be attached for `[` to dispatch
data_path <- "data/"
figure_path <- "figures/final-selected/Figure 6/"
source("code/R/gated_protein_helpers.R")
# 6a: hand-drawn schematic -- not code-generated, no output here.Panels (b) onward additionally load the shared CITE-seq cell/protein setup:
# ============================================================
# Load data (shared with FigureS6.R)
# ============================================================
source("code/R/citeseq_shared_setup.R")
# Re-derive the normalized protein matrix used only by this script's panel
# 6b (Protein_F_pm_raw filtered/scaled -- FigureS6.R doesn't need it).
Protein_F_pm <- Protein_F_pm_raw[!rownames(Protein_F_pm_raw) %in% isotype_proteins, ]
Protein_F_pm <- Protein_F_pm[rownames(Protein_F_pm) %in% good_proteins, ]
Protein_F_pm <- Protein_F_pm[!rownames(Protein_F_pm) %in% exclude_proteins, ]
Protein_F_pm <- Protein_F_pm[!rownames(Protein_F_pm) %in% thy11_proteins, ]
D_lognorm <- diag(1 / apply(Protein_F_pm, 2, function(x) max(abs(x))))
Protein_F_pm <- Protein_F_pm %*% D_lognorm
colnames(Protein_F_pm) <- paste0("GP", 1:ncol(Protein_F_pm))
Protein_F_pm[is.na(Protein_F_pm)] <- 0(a) Protein-signature schematic
This panel is a hand-drawn schematic, not generated from R – there is
no code or pre-rendered image to show here. See
figures/Previous/bits/Figure 6/6a.pdf for the published
panel.
Fig. 6a. Schematic illustrating how protein signatures were defined for each GP. Holding the cell-loading matrix L fixed from the GP model, a protein factor matrix was estimated by EBMF from the paired CITE-seq protein measurements, Y ≈ LUᵀ (cells × proteins matrix), yielding a protein signature U (proteins × GPs matrix) for each GP. See Methods: fitting CITE-seq protein programs with fixed scRNA loadings.
(b) Protein-program heatmap
# ============================================================
# 6b: sparse protein-program heatmap, contamination GPs removed
# ============================================================
threshold_simplified <- 0
keep_rows_simplified <- apply(Protein_F_pm, 1, function(v) any(abs(v) > threshold_simplified, na.rm = TRUE))
Protein_F_pm_simplified <- Protein_F_pm[keep_rows_simplified, , drop = FALSE]
keep_cols_simplified <- apply(Protein_F_pm_simplified, 2, function(v) any(abs(v) > threshold_simplified, na.rm = TRUE))
Protein_F_pm_simplified <- Protein_F_pm_simplified[, keep_cols_simplified, drop = FALSE]
GP_contamination <- c("GP40", "GP50", "GP55", "GP188")
Protein_F_pm_simplified_no_contamination <- Protein_F_pm_simplified[, !colnames(Protein_F_pm_simplified) %in% GP_contamination, drop = FALSE]
sparse_cutoff <- 0.5
bk_sparse <- unique(c(seq(-1, -sparse_cutoff, length.out = 26), seq(-sparse_cutoff, sparse_cutoff, length.out = 51), seq(sparse_cutoff, 1, length.out = 26)))
cols_sparse <- c(colorRampPalette(c("#4575B4", "white"))(25), rep("white", 50), colorRampPalette(c("white", "#D73027"))(25))
# Display proteins as rows and GPs as columns. Order GP columns from most to
# fewest visible proteins. Order protein rows by their rightmost visible GP, so
# proteins extending into the sparse right side appear first and form a
# triangular boundary. Visible count and a rarity-weighted support score provide
# deterministic secondary ordering.
wide_matrix_6b <- as.matrix(Protein_F_pm_simplified_no_contamination)
wide_visible_mask_6b <- abs(wide_matrix_6b) >= sparse_cutoff
wide_gp_visible_count_6b <- colSums(wide_visible_mask_6b)
wide_protein_visible_count_6b <- rowSums(wide_visible_mask_6b)
wide_gp_number_6b <- as.integer(sub("^GP", "", colnames(wide_matrix_6b)))
wide_gp_order_6b <- order(-wide_gp_visible_count_6b, wide_gp_number_6b)
wide_mask_ordered_cols_6b <- wide_visible_mask_6b[
,
wide_gp_order_6b,
drop = FALSE
]
wide_rightmost_visible_gp_6b <- apply(
wide_mask_ordered_cols_6b,
1,
function(values) max(which(values))
)
wide_rarity_weights_6b <- seq_len(ncol(wide_mask_ordered_cols_6b))^2
wide_protein_rarity_score_6b <- as.numeric(
wide_mask_ordered_cols_6b %*% wide_rarity_weights_6b
)
wide_protein_order_6b <- order(
-wide_rightmost_visible_gp_6b,
-wide_protein_visible_count_6b,
-wide_protein_rarity_score_6b,
rownames(wide_matrix_6b)
)
wide_ordered_matrix_6b <- wide_matrix_6b[
wide_protein_order_6b,
wide_gp_order_6b,
drop = FALSE
]
pdf(paste0(figure_path, "6b.pdf"), width = 48, height = 14)
pheatmap::pheatmap(
wide_ordered_matrix_6b,
main = sprintf(
paste0(
"Protein programs - GP columns, triangular-first protein rows ",
"(|score| >= %.1f; protein-row sparsity is not monotone)"
),
sparse_cutoff
),
color = cols_sparse,
breaks = bk_sparse,
cluster_rows = FALSE,
cluster_cols = FALSE,
border_color = "grey75",
fontsize = 16,
fontsize_row = 16,
fontsize_col = 16,
angle_col = 90,
legend_breaks = c(-1, -sparse_cutoff, 0, sparse_cutoff, 1),
legend_labels = c("-1", "-0.5", "0 (white)", "0.5", "1")
)
dev.off()
Fig. 6b. Heatmap of scaled protein scores for each GP (the re-estimated protein matrix U), with the protein scores in each GP scaled so its maximum |score| = 1. We focused on the 47 proteins that performed best in the immgenT CITE-seq dataset (isotype and low-quality proteins removed). Seventeen GPs lacked meaningful protein correlates – every protein score is exactly zero – and were omitted. Four putative contamination programs (GP40, GP50, GP55 and GP188) were also removed, leaving 179 GPs (200 - 17 - 4 = 179); 47 proteins (rows) x 179 GPs (columns). Entries with an absolute score below 0.5 are shown in white, with color running from blue (-1) through white to red (+1). GP columns are ordered from the most to the fewest visible proteins. Protein rows are ordered by the position of their rightmost visible GP, placing proteins that extend into the sparse right side first and exposing a triangular support boundary; visible count and a rarity-weighted support score break ties. This triangular-first row order is not monotone in per-protein sparsity. GP labels are vertical and sized to the available heatmap-cell width; protein labels are sized to the available row height.
(c-f) Protein-gated vs. GP-loading populations
# ============================================================
# 6c-6f: protein-gate vs. GP-loading comparison for the 4 curated main-figure GPs
# (df_markers2, thymocyte/proliferating/miniverse_cells, L_pm_for_gating,
# select_proteins, threshold_results_subset_manual all come from
# citeseq_shared_setup.R above)
# ============================================================
# Panel lettering follows the published figure: c = GP171, d = GP12, e = GP80,
# f = GP23. (It is NOT the order the GPs happen to be listed in below -- an
# earlier version of this script assigned the letters positionally, which
# silently swapped d/e/f relative to the published panels.)
GPs_fig6 <- c("GP171", "GP23", "GP12", "GP80")
fig6_letter <- c("GP171" = "6c", "GP12" = "6d", "GP80" = "6e", "GP23" = "6f")
enlarge_gps <- c("GP8", "GP30", "GP170", "GP107")
for (gp in GPs_fig6) {
k_name <- paste0("K", sub("^GP", "", gp))
plot_gated_gp_vs_protein(
gp_name = k_name,
df_markers = df_markers2,
protein_mat = protein_mat_normalized_lognorm,
loading_mat = L_pm_for_gating,
mde_emb = mde_result,
missing_threshold_action = "skip",
threshold_df = threshold_results_subset_manual,
exclude_cells = c(thymocyte_cells, proliferating_cells, miniverse_cells),
selected_proteins = select_proteins,
loading_q = NULL,
min_pointsize = if (gp %in% enlarge_gps) 3L else 0L,
save_path = paste0(figure_path, fig6_letter[gp], ".pdf")
)
}
Fig. 6c-f. Examples of gating strategies used to identify GP-active cells, for (c) GP171, (d) GP12, (e) GP80 and (f) GP23. For each GP, cells are shown twice on the same MDE embedding. Left, cells selected by the proposed protein gate, built from the GP’s curated marker signature (positive markers above, and negative markers below); right, an equally sized set of cells with the highest GP activity (the loading cutoff is chosen to match the protein-gate count). Color indicates cell density (two-dimensional); all other cells are grey. Thymocytes, proliferating, and “miniverse” cells are excluded.
(g, h) KLRG1 modulation across lineages
# ============================================================
# 6g/6h: KLRG1 modulation (CD8 vs CD4, CD8 vs Treg)
# ============================================================
FlashierDGE_corrected <- function(F1, L1, group1, group2, title_plot = "") {
loadings_group1 <- colMeans(L1[group1, ])
loadings_group2 <- colMeans(L1[group2, ])
mean_change_loadings <- loadings_group1 - loadings_group2
vplot <- data.frame(SYMBOL = names(mean_change_loadings), mean_change_loadings = mean_change_loadings, AveExpr = colMeans(L1[c(group1, group2), ]))
list(diff_factors = vplot)
}
get_klrg1_split <- function(cell_type_label, meta, protein_data, threshold) {
cells <- meta$cellID[meta$annotation_level1 == cell_type_label]
cells <- intersect(cells, rownames(protein_data))
list(pos = cells[protein_data[cells, "KLRG1"] >= threshold], neg = cells[protein_data[cells, "KLRG1"] < threshold])
}
run_checked_dge <- function(group_list, F_mat, L_mat, label) {
if (length(group_list$pos) < 3 || length(group_list$neg) < 3) stop(paste("Insufficient data:", label))
df <- FlashierDGE_corrected(F1 = F_mat, L1 = L_mat, group1 = group_list$pos, group2 = group_list$neg)$diff_factors
if (!"SYMBOL" %in% colnames(df)) df$SYMBOL <- rownames(df)
df
}
plot_target_gps <- function(df, x_var, y_var, label_var, target_gps, highlight_color = "darkorange", background_color = "black",
x_limits = c(-0.5, 0.5), y_limits = c(-0.5, 0.5), background_alpha = 0.5,
xlab = "Difference in Mean Loading", ylab = "Difference in Mean Loading", title = "Comparison of Specific GP Loadings") {
highlight_df <- df %>% filter({{ label_var }} %in% target_gps) %>% mutate(.label_display = as.character({{ label_var }}))
ggplot(df, aes(x = {{ x_var }}, y = {{ y_var }})) +
geom_point(color = background_color, alpha = background_alpha) +
geom_abline(slope = 1, intercept = 0, linetype = "dashed", color = "red") +
geom_hline(yintercept = 0, linetype = "dashed", color = "blue") +
geom_vline(xintercept = 0, linetype = "dashed", color = "blue") +
geom_point(data = highlight_df, aes(color = {{ label_var }}), size = 2) +
ggrepel::geom_text_repel(seed = 42, data = highlight_df, aes(label = .label_display, color = {{ label_var }}), max.overlaps = Inf, size = 3.5, box.padding = 0.35, point.padding = 0.5, segment.color = "grey50", show.legend = FALSE) +
scale_color_manual(values = highlight_color, guide = "none") +
coord_cartesian(xlim = x_limits, ylim = y_limits) +
labs(x = xlab, y = ylab, title = title) +
theme_minimal()
}
klrg1_threshold <- threshold_results_subset_manual$Threshold[threshold_results_subset_manual$Protein == "KLRG1"]
cd8_split <- get_klrg1_split("CD8", seurat_meta_filtered, protein_mat_normalized_lognorm, klrg1_threshold)
CD4_split <- get_klrg1_split("CD4", seurat_meta_filtered, protein_mat_normalized_lognorm, klrg1_threshold)
treg_split <- get_klrg1_split("Treg", seurat_meta_filtered, protein_mat_normalized_lognorm, klrg1_threshold)
diff_CD8 <- run_checked_dge(cd8_split, F_pm_filtered, L_pm_filtered, "CD8") %>% rename(mean_change_CD8 = mean_change_loadings, AveExpr_CD8 = AveExpr)
diff_CD4 <- run_checked_dge(CD4_split, F_pm_filtered, L_pm_filtered, "CD4") %>% rename(mean_change_CD4 = mean_change_loadings, AveExpr_CD4 = AveExpr)
diff_Treg <- run_checked_dge(treg_split, F_pm_filtered, L_pm_filtered, "Treg") %>% rename(mean_change_Treg = mean_change_loadings, AveExpr_Treg = AveExpr)
# 6g: CD8 vs CD4
merged_cd4 <- inner_join(diff_CD4, diff_CD8, by = "SYMBOL")
p_6g <- plot_target_gps(
df = merged_cd4, x_var = mean_change_CD8, y_var = mean_change_CD4, label_var = SYMBOL,
target_gps = c("GP10", "GP58", "GP25", "GP26", "GP43"), background_alpha = 0.8, x_limits = c(-0.2, 0.4), y_limits = c(-0.2, 0.4),
highlight_color = c("GP10" = "darkorange2", "GP25" = "blue", "GP43" = "blue", "GP26" = "blue", "GP58" = "darkorange2"),
title = "KLRG1 Modulation: CD8 vs CD4", xlab = "Effect Size in CD8 (KLRG1+ - KLRG1-)", ylab = "Effect Size in CD4 (KLRG1+ - KLRG1-)"
) + theme_bw()
ggsave(paste0(figure_path, "6g.pdf"), p_6g, width = 7, height = 6)
# 6h: CD8 vs Treg
merged_treg <- inner_join(diff_Treg, diff_CD8, by = "SYMBOL")
p_6h <- plot_target_gps(
df = merged_treg, x_var = mean_change_CD8, y_var = mean_change_Treg, label_var = SYMBOL,
target_gps = c("GP6", "GP10", "GP12", "GP27", "GP68", "GP58"), background_alpha = 0.8, x_limits = c(-0.2, 0.4), y_limits = c(-0.2, 0.4),
highlight_color = c("GP10" = "darkorange2", "GP27" = "deeppink", "GP6" = "deeppink", "GP68" = "deeppink", "GP12" = "deeppink", "GP58" = "darkorange2"),
title = "KLRG1 Modulation: CD8 vs Treg", xlab = "Effect Size in CD8 (KLRG1+ - KLRG1-)", ylab = "Effect Size in Treg (KLRG1+ - KLRG1-)"
) + theme_bw()
ggsave(paste0(figure_path, "6h.pdf"), p_6h, width = 7, height = 6)
Fig. 6g, h. GPs associated with KLRG1 protein expression in CD4, Tregs and CD8. Within each lineage, cells are split KLRG1+ versus KLRG1- based on the threshold on the KLRG1 protein, and for every GP the effect size is the difference in its mean activity between KLRG1+ and KLRG1- cells. These effect-size versus effect-size plots compare CD8+ T cells (x-axis) with CD4+ T cells (g) or Treg cells (h) (y-axis); each point is a GP, the dashed red line marks equal effect (y = x), and the dashed blue lines mark zero. Selected GPs are labeled – orange for large, concordant effects in both lineages, and blue/pink for lineage-biased GPs.
(i-k) GPs associated with CD69
# ============================================================
# 6i/6j/6k: 10 curated GPs from among those most associated with CD69
# ============================================================
D_scale6 <- diag(1 / apply(F_pm_filtered, 2, function(x) max(abs(x), na.rm = TRUE)))
F_pm_filtered_scaled <- F_pm_filtered %*% D_scale6
colnames(F_pm_filtered_scaled) <- paste0("GP", 1:ncol(F_pm_filtered_scaled))
# Curated list -- NOT a top-10 computed from the correlations below. These 10
# are drawn from among the most strongly CD69-correlated GPs, but are not the
# top 10 under any single ranking: 8 are positively correlated (ranks 1, 3, 4,
# 5, 6, 10, 12, 14 of 200) and GP58/GP171 are the two most *negatively*
# correlated GPs of all 200. By |rho| they sit at ranks 1, 2, 4, 5, 6, 8, 12,
# 14, 17, 18, skipping GP1/GP47/GP100/GP25. Treat as a hand-picked input like
# Thresholds_Selected_Proteins.csv and well_aligned_gps -- don't "fix" it into
# a computed ranking, and keep the caption's "from among the most associated"
# wording in sync (analysis/Figure6.Rmd, Fig. 6i-k).
cd69_top_gps_subset <- c("GP35", "GP6", "GP170", "GP26", "GP58", "GP171", "GP63", "GP62", "GP3", "GP29")
shared_cells_cd69 <- intersect(rownames(L_pm_filtered), rownames(protein_mat_normalized_lognorm))
cd69_expr_vec <- protein_mat_normalized_lognorm[shared_cells_cd69, "CD69"]
cd69_corr <- sapply(cd69_top_gps_subset, function(gp) cor(L_pm_filtered[shared_cells_cd69, gp], cd69_expr_vec, method = "spearman"))
cd69_top_gps_sorted <- names(sort(cd69_corr, decreasing = FALSE)) # most-correlated GP ends up at top of y-axis
plot_factor_heatmap <- function(F_matrix, gp_vector, n_top = 5, min_abs_loading = 0.5, transpose = FALSE,
title = "Factor loadings – top genes per GP", low_color = "steelblue", mid_color = "white", high_color = "firebrick", font_size = 9) {
F_sub <- F_matrix[, gp_vector, drop = FALSE]
selected_genes <- lapply(gp_vector, function(gp) {
vals <- F_sub[, gp]
top_pos <- names(sort(vals, decreasing = TRUE))[seq_len(min(n_top, sum(vals > 0)))]
top_neg <- names(sort(vals, decreasing = FALSE))[seq_len(min(n_top, sum(vals < 0)))]
c(top_pos, top_neg)
})
selected_genes <- unique(unlist(selected_genes))
if (min_abs_loading > 0) {
max_abs <- apply(F_sub[selected_genes, , drop = FALSE], 1, function(x) max(abs(x), na.rm = TRUE))
selected_genes <- names(max_abs[max_abs >= min_abs_loading])
}
hc_genes <- hclust(dist(F_sub[selected_genes, , drop = FALSE]))
gene_order <- rownames(F_sub[selected_genes, , drop = FALSE])[hc_genes$order]
plot_df <- F_sub[selected_genes, , drop = FALSE] %>%
as.data.frame() %>%
tibble::rownames_to_column("Gene") %>%
tidyr::pivot_longer(cols = -Gene, names_to = "GP", values_to = "Loading") %>%
mutate(GP = factor(GP, levels = gp_vector), Gene = factor(Gene, levels = gene_order))
limit <- max(abs(plot_df$Loading), na.rm = TRUE)
x_aes <- if (transpose) "Gene" else "GP"
y_aes <- if (transpose) "GP" else "Gene"
ggplot(plot_df, aes(x = .data[[x_aes]], y = .data[[y_aes]], fill = Loading)) +
geom_tile() +
scale_fill_gradient2(low = low_color, mid = mid_color, high = high_color, midpoint = 0, limits = c(-limit, limit), name = "Loading") +
# Axis titles are the faceting variables themselves ("Gene" / "GP"), as in
# the published 6i -- an earlier version dropped them along with the title's
# "- top genes per GP" suffix.
labs(title = title, x = x_aes, y = y_aes) +
theme_minimal(base_size = font_size) +
theme(axis.text.x = element_text(angle = if (transpose) 90 else 45, hjust = 1, size = font_size), axis.text.y = element_text(size = font_size), panel.grid = element_blank(), plot.title = element_text(face = "bold"))
}
p_heatmap <- plot_factor_heatmap(F_matrix = F_pm_filtered_scaled, gp_vector = cd69_top_gps_sorted, n_top = 5, font_size = 9, transpose = TRUE, low_color = "#4DAF4A", mid_color = "white", high_color = "#984EA3")
corr_strip_df <- data.frame(GP = factor(cd69_top_gps_sorted, levels = cd69_top_gps_sorted), Correlation = cd69_corr[cd69_top_gps_sorted], x = "Corr")
corr_limit <- max(abs(corr_strip_df$Correlation))
p_corr_strip <- ggplot(corr_strip_df, aes(x = x, y = GP, fill = Correlation)) +
geom_tile() +
scale_fill_gradient2(low = "royalblue", mid = "white", high = "tomato", midpoint = 0, limits = c(-corr_limit, corr_limit), name = "Corr\n(CD69)") +
labs(x = NULL, y = NULL) +
theme_minimal(base_size = 9) +
theme(axis.text.x = element_text(size = 9, angle = 45, hjust = 1), axis.text.y = element_blank(), axis.ticks.y = element_blank(), panel.grid = element_blank())
p_6i <- p_corr_strip + p_heatmap + patchwork::plot_layout(widths = c(0.06, 1), guides = "collect")
ggsave(paste0(figure_path, "6i.pdf"), p_6i, width = 11, height = 5)
# 6j/6k: mean loading of these GPs per tissue (j) and per lineage (k)
cells_for_heatmap <- intersect(rownames(L_pm_filtered), rownames(seurat_meta_filtered))
L_cd69_sub <- L_pm_filtered[cells_for_heatmap, cd69_top_gps_sorted, drop = FALSE]
meta_hm <- seurat_meta_filtered[cells_for_heatmap, c("annotation_level1", "organ_simplified")]
mean_loading_long <- function(L_mat, group_vec, gp_levels) {
as.data.frame(L_mat) %>%
mutate(group = group_vec) %>%
tidyr::pivot_longer(cols = -group, names_to = "GP", values_to = "Loading") %>%
group_by(group, GP) %>%
summarise(mean_loading = mean(Loading, na.rm = TRUE), .groups = "drop") %>%
mutate(GP = factor(GP, levels = gp_levels))
}
make_mean_loading_heatmap <- function(df, title) {
fill_max <- max(df$mean_loading, na.rm = TRUE)
ggplot(df, aes(x = group, y = GP, fill = mean_loading)) +
geom_tile() +
scale_fill_gradient(low = "white", high = "firebrick", limits = c(0, fill_max), name = "Mean\nloading") +
labs(title = title, x = NULL, y = NULL) +
theme_minimal(base_size = 9) +
theme(axis.text.x = element_text(angle = 45, hjust = 1, size = 9), axis.text.y = element_text(size = 9), panel.grid = element_blank())
}
df_organ <- mean_loading_long(L_cd69_sub, meta_hm$organ_simplified, cd69_top_gps_sorted)
p_6j <- make_mean_loading_heatmap(df_organ, "Mean GP loading by tissue (organ_simplified)")
ggsave(paste0(figure_path, "6j.pdf"), p_6j, width = 9, height = 5)
df_level1 <- mean_loading_long(L_cd69_sub, meta_hm$annotation_level1, cd69_top_gps_sorted)
p_6k <- make_mean_loading_heatmap(df_level1, "Mean GP loading by cell type (level1)")
ggsave(paste0(figure_path, "6k.pdf"), p_6k, width = 7, height = 5)
Fig. 6i-k. GPs associated with CD69 protein expression. Ten representative GPs associated with CD69, selected from among those most strongly correlated – positively or negatively – with CD69 protein expression, ordered by the Spearman correlation between GP activity and CD69 expression. (i) Heatmap of the genes most strongly up- and down-regulated across these GPs: per GP, the top five positively and top five negatively scoring genes, then restricted to genes reaching |score| >= 0.5 in at least one of the ten (green-white-purple), with the left strip indicating each GP’s Spearman correlation with CD69 (blue-white-red). (j) Mean activity of these GPs across tissues and (k) across lineages.
The ten GPs are a hand-picked set, not a computed
top-10: eight are positively correlated with CD69 (ranks 1, 3, 4, 5, 6,
10, 12 and 14 of 200) and GP58/GP171 are the two most
negatively correlated GPs of all 200.
script/verify_cd69_gp_ranking.R recomputes the correlation
over all 200 GPs and fails if this wording and the code drift apart.
sessionInfo()R version 4.5.1 (2025-06-13)
Platform: aarch64-apple-darwin20
Running under: macOS Sequoia 15.6.1
Matrix products: default
BLAS: /System/Library/Frameworks/Accelerate.framework/Versions/A/Frameworks/vecLib.framework/Versions/A/libBLAS.dylib
LAPACK: /Library/Frameworks/R.framework/Versions/4.5-arm64/Resources/lib/libRlapack.dylib; LAPACK version 3.12.1
locale:
[1] en_CA/en_CA/en_CA/C/en_CA/en_CA
time zone: America/Chicago
tzcode source: internal
attached base packages:
[1] stats graphics grDevices utils datasets methods base
loaded via a namespace (and not attached):
[1] vctrs_0.7.3 cli_3.6.6 knitr_1.50 rlang_1.2.0
[5] xfun_0.55 stringi_1.8.7 otel_0.2.0 promises_1.5.0
[9] jsonlite_2.0.0 workflowr_1.7.2 glue_1.8.1 rprojroot_2.1.1
[13] git2r_0.36.2 htmltools_0.5.9 httpuv_1.6.16 sass_0.4.10
[17] rmarkdown_2.30 evaluate_1.0.5 jquerylib_0.1.4 tibble_3.3.0
[21] fastmap_1.2.0 yaml_2.3.12 lifecycle_1.0.5 whisker_0.4.1
[25] stringr_1.6.0 compiler_4.5.1 fs_1.6.6 Rcpp_1.1.1-1.1
[29] pkgconfig_2.0.3 later_1.4.4 digest_0.6.39 R6_2.6.1
[33] pillar_1.11.1 magrittr_2.0.5 bslib_0.9.0 tools_4.5.1
[37] cachem_1.1.0