Which method apply to UCSC Xena RNASeq data for DE anlaysis

[ShixiangWang]

Suggestion from @ucscXena,

• (1 of 3) For DESeq2 and EdgeR, the values in the matrix should be un-normalized counts or estimated counts of sequencing reads (for single-end RNA-seq) or fragments (for paired-end RNA-seq).
• (2 of 3) If the data download from Xena are in the unit of log2(count +1) or log2(expected_count+1), you can use the datasets, but you need to reverse the log-transformation first before using them in DESeq2 or EdgeR.
• (3 of 3) For example, RSEM or Kallisto generates results in unit of expected_count, and HTSeq in the unit of count. For Limma package on RNAseq data, it also needs count matrix, so the same above applies.

Really thanks and I have some problem when I reversed the log-transformation, I found the reversed count matrix are not integer values, could I apply this matrix to DESeq2 or EdgeR? I tried 2 based and exp based log transformation.

I use LUAD data as an example, data link

> library(UCSCXenaTools)
> luad_get = UCSCXenaTools::getTCGAdata(project = "LUAD", clinical = F, mRNASeq = TRUE, download = TRUE)
We will download files to directory /var/folders/mx/rfkl27z90c96wbmn3_kjk8c80000gn/T//RtmpiOg0Fu.
Downloading TCGA.LUAD.sampleMap__HiSeqV2.gz
试开URL’https://tcga.xenahubs.net/download/TCGA.LUAD.sampleMap/HiSeqV2.gz'
Content type 'application/gzip' length 30884695 bytes (29.5 MB)
==================================================
downloaded 29.5 MB

Note fileNames transfromed from datasets name and / chracter all changed to __ character.
> luad_rna = UCSCXenaTools::XenaPrepare(luad_get)
> luad_rna[1:5, 1:5]
# A tibble: 5 x 5
  sample    `TCGA-69-7978-01` `TCGA-62-8399-01` `TCGA-78-7539-01` `TCGA-50-5931-11`
  <chr>                 <dbl>             <dbl>             <dbl>             <dbl>
1 ARHGEF10L             9.99               10.4              9.63             8.68 
2 HIF3A                 4.26               11.6              9.14             9.48 
3 RNF17                 0.418               0                1.12             0.822
4 RNF10                10.4                11.5             11.7             11.7  
5 RNF11                11.2                11.0             10.5             11.7  
> small_luad = luad_rna[1:100, 1:10]
> luad_raw_count = 2^small_luad[,-1] - 1
> luad_raw_count
    TCGA-69-7978-01 TCGA-62-8399-01 TCGA-78-7539-01 TCGA-50-5931-11 TCGA-73-4658-01 TCGA-44-6775-01
1      1.015786e+03    1.374461e+03    7.893786e+02    4.101440e+02    5.903219e+02     1025.771897
2      1.815700e+01    3.155040e+03    5.616914e+02    7.142977e+02    3.164522e+01       15.648266
3      3.361667e-01    0.000000e+00    1.178145e+00    7.679776e-01    0.000000e+00        1.009588
4      1.318431e+03    2.995162e+03    3.255711e+03    3.405560e+03    3.148485e+03     2306.476116
5      2.305996e+03    2.075589e+03    1.415289e+03    3.277227e+03    2.593786e+03     2189.280612
6      1.540052e+03    6.225234e+02    1.412347e+03    2.979214e+03    1.960307e+03     1955.012670

    TCGA-44-2655-01 TCGA-44-3398-01 TCGA-62-8397-01
1      6.409424e+02    5.199135e+02     1489.003496
2      4.531797e+01    5.224964e+01       99.266788
3      5.179245e-01    0.000000e+00        0.000000
4      3.865735e+03    2.568728e+03     5259.922200
5      2.709607e+03    2.890295e+03     2490.501971
6      2.374969e+03    3.575915e+03      908.040020

> luad_raw_count = exp(small_luad[,-1]) - 1
> head(luad_raw_count)
  TCGA-69-7978-01 TCGA-62-8399-01 TCGA-78-7539-01 TCGA-50-5931-11 TCGA-73-4658-01 TCGA-44-6775-01 TCGA-44-2655-01
1    2.180194e+04        33714.06     15158.76261    5.903677e+03       9973.6285    22111.536741    1.122851e+04
2    6.979582e+01       111735.65      9284.41344    1.312565e+04        151.7496       56.817992    2.520280e+02
3    5.190726e-01            0.00         2.07437    1.275273e+00          0.0000        1.737103    8.259493e-01
4    3.175065e+04       103661.94    116913.71309    1.247531e+05     111400.9377    71117.302373    1.497756e+05
5    7.109597e+04        61082.68     35167.28637    1.180297e+05      84236.8817    65965.347196    8.971460e+04
6    3.972257e+04        10766.63     35061.93961    1.028668e+05      56251.1758    56032.219545    7.418272e+04
  TCGA-44-3398-01 TCGA-62-8397-01
1       8306.3837       37838.167
2        308.4201         770.009
3          0.0000           0.000
4      83065.7682      233536.980
5      98468.2898       79442.392
6     133852.3397       18548.535

Another problem is for RNASeq data type is pancan normalization or percentile, they are hard to get count matrix, am I right? But all three RNAseq data type I mentioned basically follow normal distribution, I think I can treat them as microarray data and all use limma package (not voom method) to get DE genes. Does this idea work?

Shixiang