✦
医学科研新动向
✦
TCGA
TCGA（The Cancer Genome Atlas）是一个拥有海量肿瘤基因组数据的公共数据库，为癌症研究提供了丰富的资源。掌握如何高效获取并分析这些数据对研究至关重要！
📊 本篇推文将快速了解：如何从TCGA获取肿瘤miRNA基因表达数据。
登录TCGA网站来下载数据：https://portal.gdc.cancer.gov/
/ TO MOM /
/ TO MOM /
01
数据下载
01. 点击 Cohort Builder
02. 以TCGA-DLBC为例 → Program → TCGA → Project → TCGA-DLBC
03. 点击Repository
04. 下载必须文件
Metadata + Clinical.csv
02
miRNA表达矩阵获取
将下载的文件保存在分析路径中：
# 设置工作目录setwd("C:\\Users\\admin\\Desktopdata\\miRNA")# 安装包if (!requireNamespace("BiocManager", quietly = TRUE))install.packages("BiocManager")BiocManager::install("edgeR")BiocManager::install("miRBaseVersions.db")# 加载必要的库library(rjson)library(edgeR)library(data.table)library(dplyr)library(stringr)library(tidyr)library(miRBaseVersions.db)# 读取 metadata.json 文件json <- jsonlite::fromJSON("metadata.cart.2024-10-22.json")file_sample0 <- json[c('file_name', 'associated_entities')]# 提取样本IDfile_sample0$sample_id <- sapply(file_sample0$associated_entities, function(x) { x[, 1] })file_sample <- subset(file_sample0, select = -associated_entities)# 列出所有miRNA的表达文件count_file <- list.files('gdc_download_20241022_072814.603834', pattern = '*.isoforms.quantification.tsv', recursive = TRUE)count_file_name <- sapply(strsplit(count_file, split = '/'), function(x) { x[2] })# 创建一个空的表达矩阵isoform_matrix <- data.frame()path1 <- "gdc_download_20241022_072814.603834\\"# 循环读取每个miRNA的表达文件，并合并成表达矩阵for (i in 1:length(count_file_name)) {path <- paste0(path1, count_file[i])data0 <- read.table(path, fill = TRUE, header = TRUE)# 提取感兴趣的列：miRNA区域和 read_countdata1 <- data0[, c(6, 3)] # 列6是miRNA区域，列3是 read_countdata <- data1 %>%group_by(miRNA_region) %>%summarise_all(sum) # 合并同一miRNA区域的表达量（求和）# 设置列名为样本IDcolnames(data)[2] <- file_sample0[which(file_sample0$file_name == count_file_name[i]), 'sample_id']# 合并到总的表达矩阵isoform_matrix <- if (nrow(isoform_matrix) == 0) data else merge(isoform_matrix, data, by = "miRNA_region", all = TRUE)}# 查看生成的表达矩阵View(isoform_matrix)dim(isoform_matrix)# 导出原始表达矩阵write.table(isoform_matrix, "raw_mir_counts.txt", sep = "\t", row.names = TRUE, col.names = NA, quote = FALSE)### 数据处理，提取成熟miRNA区域isoform_matrix <- isoform_matrix[grep("mature", isoform_matrix$miRNA_region), ]exp <- isoform_matrix# 分割 miRNA_region 列exp <- separate(exp, into = c("State", "ACCESSION"), col = "miRNA_region", sep = ",")# 读取 miRBase 数据库文件ano <- read.table("miRbase.txt", sep = "\t", header = TRUE, stringsAsFactors = FALSE)merge_mir <- merge(ano, exp, by = "ACCESSION")merge_mir <- merge_mir[, -c(1, 3)] # 去除无关列# 保存过滤后的数据write.table(merge_mir, "raw_mir_counts_filtered.txt", sep = "\t", row.names = TRUE, col.names = NA, quote = FALSE)### 过滤表达量# 将表达量为 NA 的地方替换成 0merge_mir[, -1][is.na(merge_mir[, -1])] <- 0# 去除在 50% 样本中没有表达的基因zero_percentage <- rowMeans(merge_mir[, 2:ncol(merge_mir)] == 0)exp1 <- merge_mir[zero_percentage < 0.5, ]# 对重复的基因名进行平均处理selected_cols <- exp1[, -1]exp2 <- avereps(selected_cols, ID = exp1$NAME)# 去除低表达基因exp2 <- exp2[rowMeans(exp2) > 0.5, ]### 样本分组# 按样本编号提取肿瘤和正常样本tumor <- colnames(exp2)[as.integer(substr(colnames(exp2), 14, 15)) == 01]normal <- colnames(exp2)[as.integer(substr(colnames(exp2), 14, 15)) == 11]exprSet_by_group <- cbind(exp2[, tumor], exp2[, normal])# 修改列名，提取前16位字符colnames(exprSet_by_group) <- substr(colnames(exprSet_by_group), 1, 16)# 保存最终的表达矩阵write.table(as.data.frame(exprSet_by_group), "mir_counts_final.txt", sep = "\t", row.names = TRUE, col.names = NA, quote = FALSE)
-END-
文字丨本人编写，如有补充，请随时告诉我
后台私信TCGA-miRNA即可获取完整代码
