# This R script will take # 1) an expression matrix, # 2) a Silhouette Clusters Function file, # 3) list of differentially expressed genes along with stats and # 4) GO enrichment stats to generate a CSV file with list of functional groups and representative genes. # =================================================================== # Cluster Summary Script with GO Enrichment Integration # =================================================================== # Load libraries library(dplyr) library(readxl) library(cluster) library(openxlsx) library(stringr) # ------------------------------------------------------------------- # Function: summarize_clusters_with_GO (multi-DEG version, strict reps) # ------------------------------------------------------------------- summarize_clusters_with_GO <- function( expr_mat, # numeric matrix/array: rows=genes, cols=samples annot_xlsx, # Excel with Gene_ID + Silhouette_Cluster (+ Function if available) deg_files, # character vector: one or more DEG .tsv files go_folder = "GO_files", # folder containing GO_Enrichment_Cluster_X.csv figure_label = "LeafBlue+LeafRed; Common DEGs", blue_pattern = "Root_Blue", # regex for blue samples red_pattern = "Root_Red", # regex for red samples sentinel_floor = -3.231646, n_funcs = 3, n_reps = 3 ){ stopifnot(is.matrix(expr_mat) || is.array(expr_mat)) expr <- expr_mat # --- Clean expression expr[expr == sentinel_floor] <- NA expr <- expr[rowSums(!is.na(expr)) > 0, , drop = FALSE] expr <- expr[complete.cases(expr), , drop = FALSE] expr_scaled <- scale(expr) # --- Detect Blue vs Red columns blue_cols <- grep(blue_pattern, colnames(expr_scaled), value = TRUE) red_cols <- grep(red_pattern, colnames(expr_scaled), value = TRUE) if (length(blue_cols) == 0 || length(red_cols) == 0) { stop("Could not find Blue or Red columns. Check blue_pattern/red_pattern.") } # --- Read annotation annot <- readxl::read_excel(annot_xlsx) id_col <- intersect(names(annot), c("Gene_ID","Gene","transcript_id","Gene stable ID"))[1] names(annot)[names(annot) == id_col] <- "Gene_ID" clust_col <- intersect(names(annot), c("Silhouette_Cluster","Cluster"))[1] names(annot)[names(annot) == clust_col] <- "Silhouette_Cluster" func_col <- intersect(names(annot), c("Function","Annotation","Description"))[1] if (!is.na(func_col)) { names(annot)[names(annot) == func_col] <- "Function" } else { annot$Function <- NA_character_ } annot <- annot %>% dplyr::select(Gene_ID, Silhouette_Cluster, Function) # --- Read and intersect multiple DEG tables deg_list <- lapply(deg_files, function(f) { d <- read.delim(f, stringsAsFactors = FALSE) if (!"Gene_ID" %in% names(d)) names(d)[1] <- "Gene_ID" if (!"log2FC" %in% names(d)) d$log2FC <- NA_real_ if (!"FDR" %in% names(d)) d$FDR <- NA_real_ d }) # find common genes across all DEG files common_genes <- Reduce(intersect, lapply(deg_list, function(x) x$Gene_ID)) # restrict each DEG table to common genes deg_list <- lapply(seq_along(deg_list), function(i) { d <- deg_list[[i]] d <- d[d$Gene_ID %in% common_genes, ] names(d)[names(d) == "log2FC"] <- paste0("log2FC_", i) names(d)[names(d) == "FDR"] <- paste0("FDR_", i) d }) # merge all into one big DEG table deg <- Reduce(function(x, y) dplyr::full_join(x, y, by="Gene_ID"), deg_list) # --- Align annotation with expr genes_in_both <- intersect(rownames(expr_scaled), annot$Gene_ID) annot_filt <- annot %>% dplyr::filter(Gene_ID %in% genes_in_both) annot_filt <- annot_filt[match(rownames(expr_scaled), annot_filt$Gene_ID), ] keep <- !is.na(annot_filt$Silhouette_Cluster) expr_scaled <- expr_scaled[keep, , drop = FALSE] annot_filt <- annot_filt[keep, , drop = FALSE] # --- Silhouette with pre-assigned cluster labels cl_vec <- as.integer(factor(annot_filt$Silhouette_Cluster)) d <- dist(expr_scaled) sil <- cluster::silhouette(cl_vec, d) sil_df <- data.frame( Gene_ID = rownames(expr_scaled), Cluster = sil[, "cluster"], Neighbor_Cluster = sil[, "neighbor"], Silhouette_Width = sil[, "sil_width"], stringsAsFactors = FALSE ) # --- Merge DEG + annotation merged <- sil_df %>% dplyr::left_join(annot_filt, by = "Gene_ID") %>% dplyr::left_join(deg, by = "Gene_ID") # --- Compute highlights (Blue vs Red mean expression) merged$mean_blue <- rowMeans(expr_scaled[, blue_cols, drop = FALSE], na.rm = TRUE) merged$mean_red <- rowMeans(expr_scaled[, red_cols, drop = FALSE], na.rm = TRUE) cluster_delta <- merged %>% dplyr::group_by(Cluster) %>% dplyr::summarize( Delta_Blue_minus_Red = mean(mean_blue - mean_red, na.rm = TRUE), IQR_within = IQR(c(mean_blue, mean_red), na.rm = TRUE), .groups = "drop" ) %>% dplyr::mutate(Cluster_Highlights = dplyr::case_when( Delta_Blue_minus_Red >= 0.25 ~ "Upregulated in Blue", Delta_Blue_minus_Red <= -0.25 ~ "Upregulated in Red", TRUE ~ ifelse(IQR_within >= 0.5, "Variable expression", "Balanced/low-shift") )) # --- Stats cluster_stats <- merged %>% dplyr::group_by(Cluster) %>% dplyr::summarize( Num_Genes = dplyr::n(), Avg_Silhouette = mean(Silhouette_Width, na.rm = TRUE), .groups = "drop" ) # --- Functional groups func_summary <- merged %>% dplyr::filter(!is.na(Function) & Function != "") %>% dplyr::count(Cluster, Function, sort = TRUE) %>% dplyr::group_by(Cluster) %>% dplyr::slice_head(n = n_funcs) %>% dplyr::summarize(Functional_Groups = paste(Function, collapse = "; "), .groups = "drop") # --- GO terms go_summary <- list() for (cl in unique(merged$Cluster)) { go_file <- file.path(go_folder, paste0("GO_Enrichment_Cluster_", cl, ".csv")) if (file.exists(go_file)) { go <- read.csv(go_file, stringsAsFactors = FALSE) if (all(c("Description","p.adjust") %in% names(go))) { top_terms <- go %>% dplyr::arrange(p.adjust) %>% head(n_funcs) go_summary[[as.character(cl)]] <- paste(top_terms$Description, collapse = "; ") } } } go_summary_df <- tibble::tibble(Cluster = as.integer(names(go_summary)), GO_Terms = unlist(go_summary)) # --- Representative genes (strict filter across all DEG files) rep_candidates <- merged %>% dplyr::filter( apply(dplyr::select(., dplyr::starts_with("log2FC_")), 1, function(x) all(abs(x) > 2, na.rm = TRUE)) & apply(dplyr::select(., dplyr::starts_with("FDR_")), 1, function(x) all(x < 0.1, na.rm = TRUE)) ) reps <- rep_candidates %>% dplyr::arrange(Cluster, dplyr::desc(Silhouette_Width)) %>% dplyr::group_by(Cluster) %>% dplyr::slice_head(n = n_reps) %>% dplyr::summarize( Representative_Gene_IDs = paste(Gene_ID, collapse = "; "), .groups = "drop" ) # --- Final table final <- cluster_stats %>% dplyr::left_join(cluster_delta %>% dplyr::select(Cluster, Cluster_Highlights), by = "Cluster") %>% dplyr::left_join(func_summary, by = "Cluster") %>% dplyr::left_join(go_summary_df, by = "Cluster") %>% dplyr::left_join(reps, by = "Cluster") %>% dplyr::mutate(Figure = figure_label, .before = 1) %>% dplyr::arrange(Cluster) # --- Save out_xlsx <- paste0(gsub("[^A-Za-z0-9]+", "_", figure_label), "_Summary_with_GO.xlsx") openxlsx::write.xlsx(list( Cluster_Table = final, PerGene_All = merged %>% dplyr::arrange(Cluster, dplyr::desc(Silhouette_Width)), Silhouette_PerGene = sil_df ), out_xlsx, row.names = FALSE) message("Wrote: ", out_xlsx) return(final) } # ------------------------------------------------------------------- # Example Usage # ------------------------------------------------------------------- # load("PDF54_edgeR_analysis.RData") final_common <- summarize_clusters_with_GO( expr_mat = heatmap_RootBlue_RootRed, annot_xlsx = "AR_Blue_AR_Red Silhouette Clusters Function.xlsx", deg_files = c("DEG AR_Blue Flt vs. Gnd.tsv", "DEG AR_Red Flt vs. Gnd.tsv"), go_folder = "AR_Blue_AR_Red", figure_label = "Common_DEGs_AR_Blue_AR_Red")