Posted on

这里使用的数据都是已经标准化去批效应的了,具体如下:

各种大型计划产出的RNA-seq数据资源已经非常丰富了,但是大家都想把多个数据库联合起来分析,就不得不面对批次效应这个问题,所以UCSC团队就使用统一的流程把这些数据重新处理了,在亚马逊云上,一个样本花费1.3美元。
发表在:Nature Biotechnology publication: https://doi.org/10.1038/nbt.3772 IF: 68.164 Q1
3大数据库是:

  1. The Cancer Genome Atlas (TCGA)
  2. Genotype-Tissue Expression (GTEx)
  3. Therapeutically Applicable Research To Generate Effective Treatments (TARGET)
    而且还提供网页工具供查询使用:

Differential gene and isoform expression of FOXM1 transcription factor in TCGA vs. GTEx

使用的数据处理流程

如下图: CutAdapt was used for adapter trimming, STAR was used for alignment, and RSEM and Kallisto were used as quantifiers.

参考基因组选择

  • STAR, RSEM, and Kallisto indexes were all built with the same reference genome. HG38 (no alt analysis) with overlapping genes from the PAR locus removed (chrY:10,000-2,781,479 and chrY:56,887,902-57,217,415).
    • ftp://ftp.ncbi.nlm.nih.gov/genomes/archive/old_genbank/Eukaryotes/vertebrates_mammals/Homo_sapiens/GRCh38/seqs_for_alignment_pipelines

注释文件的选择

  • RSEM: Gencode V23 comprehensive annotation (CHR)
    • http://www.gencodegenes.org/releases/23.html first row
  • Kallisto: Gencode V23 comprehensive annotation (ALL)
    • http://www.gencodegenes.org/releases/23.html second row

软件参数的选择

  • STAR
    • sudo docker run -v $(pwd):/data quay.io/ucsc_cgl/star –runThreadN 32 –runMode genomeGenerate –genomeDir /data/genomeDir –genomeFastaFiles hg38.fa –sjdbGTFfile gencode.v23.annotation.gtf
  • Kallisto
    • sudo docker run -v $(pwd):/data quay.io/ucsc_cgl/kallisto index -i hg38.gencodeV23.transcripts.idx transcriptome_hg38_gencodev23.fasta
    • Kallisto index that was used during the recompute is available here.
  • RSEM
    • sudo docker run -v $(pwd):/data –entrypoint=rsem-prepare-reference jvivian/rsem -p 4 –gtf gencode.v23.annotation.gtf hg38.fa hg38

Raw data

Nature Publication Supplementary Note 7 – Data Availability

Submitter sample ID to Xena sample ID mapping

TCGA mapping

GTEx mapping

TARGET mapping

最后公布的可供下载的数据集

其中TCGA TARGET GTEX 3大数据库) (共有 13 DATASETS)

cohort: TCGA TARGET GTEx

表达矩阵样本量很可观

  • RSEM expected_count
    (n=19,109)
    UCSC Toil RNAseq Recompute
  • RSEM expected_count (DESeq2 standardized)
    (n=19,039)
    UCSC Toil RNAseq Recompute
    RSEM expected_count output normalized using DESeq2
  • RSEM fpkm
    (n=19,131)
    UCSC Toil RNAseq Recompute
  • RSEM norm_count
    (n=19,120)
    UCSC Toil RNAseq Recompute
    TCGA TARGET GTEx gene expression by UCSC TOIL RNA-seq recompute
  • RSEM tpm
    (n=19,131)
    UCSC Toil RNAseq Recompute

PHENOTYPE

  • TCGA GTEX main categories
    (n=17,221)
    UCSC Toil RNAseq Recompute
  • TCGA survival data
    (n=10,496)
    UCSC Toil RNAseq Recompute
  • TCGA TARGET GTEX selected phenotypes
    (n=19,131)
    UCSC Toil RNAseq Recompute

SOMATIC MUTATION (SNP AND INDEL)

  • TCGA somatic mutations (Pan-cancer Atlas MC3 public version)
    (n=8,463)
    UCSC Toil RNAseq Recompute

TRANSCRIPT EXPRESSION RNASEQ

  • RSEM expected_count
    (n=19,109)
    UCSC Toil RNAseq Recompute
    TCGA TARGET GTEx transcript expression by RSEM using UCSC TOIL RNA-seq recompute
  • RSEM fpkm
    (n=19,129)
    UCSC Toil RNAseq Recompute
    TCGA TARGET GTEx transcript expression by RSEM using UCSC TOIL RNA-seq recompute
  • RSEM isoform percentage
    (n=19,131)
    UCSC Toil RNAseq Recompute
    TCGA TARGET GTEx transcript expression by RSEM using UCSC TOIL RNA-seq recompute
  • RSEM tpm
    (n=19,131)
    UCSC Toil RNAseq Recompute
    TCGA TARGET GTEx transcript expression by RSEM using UCSC TOIL RNA-seq recompute

6 Replies to “3大数据库超2万RNA-seq数据重新统一处理——关于TCGA-GTEx是否需要标准化”

  1. 进哥请问TCGA TARGET GTEx队列里的RSEM expected_count数据集做基因差异分析,里面的TCGA跟GTEx样本要咋处理,还需要去除批次效应然后归一化吗,归一化的话用TMM或者GeTMM可以吗

  2. 你好,进哥哥
    请问RSEM norm_count数据可否用limma包进行差异分析?SPSS等统计软件呢?

  3. 你好,进哥哥
    我想请教一下,cohort: TCGA TARGET GTEx样本表达量数据中的RSEM norm_count数据是进行了哪种处理了呢?是否有去除批次效应,是否进行了标准化?我用这个数据进行差异分析的话是否还需要进一步处理?

    1. 你好,那个文章文献里有具体讲处理流程,总的来说用的python toil工具进行的整合,已经去除batch effect,也进行了normalization;差异分析不需要再进一步对数据处理

发表评论

邮箱地址不会被公开。 必填项已用*标注