This is a test version of Biostars. For the public version, visit https://www.biostars.org.
edgeR negative bionomial error in without replicate data?

I am doing DEG analysis between two sample (Normal vs Treated) without replicates using edgeR:

library(edgeR)
data = read.table("MYFILE-counts.txt", header=T, row.names=1, com='')
bcv <- 0.2
counts <- matrix( rnbinom(40,size=1/bcv^2,mu=10), 20,2)

it is giving error:

## Warning messages:
1: In data(rnbinom(40, size = 1/bcv^2, mu = 10), 20, 2) :
 data set ‘rnbinom(40, size = 1/bcv^2, mu = 10)’ not found
2: In data(rnbinom(40, size = 1/bcv^2, mu = 10), 20, 2) :
 data set ‘20’ not found
3: In data(rnbinom(40, size = 1/bcv^2, mu = 10), 20, 2) :
 data set ‘2’ not found

I am not able to understand this error and need help to solve this. I know there is no significance of analysis without replicates, but i have no other choice.

Then i proceeded with:

counts <- data
y <- DGEList(counts=counts, group=1:2)
et <- exactTest(y, dispersion=bcv^2)

looking into the variable:

>y

An object of class "DGEList"
$counts
                     Normal.bam      Treated.bam
Cluster-0.0                  0      50
Cluster-1.0                  0      25
Cluster-2.0                  0      16
Cluster-2.1                  0       8
Cluster-3.0                  0      15


202654 more rows ...

$samples         group     lib.size      norm.factors
Normal.bam      1          3.22e+07              1
Treated.bam     2          1.05e+08              1

why there is norm.factor of 1 in both sample?

>et

    An object of class "DGEExact"
$table
            logFC logCPM   PValue
Cluster-0.0  8.03  -1.58 4.06e-05
Cluster-1.0  7.04  -2.39 2.13e-03
Cluster-2.0  6.40  -2.89 3.49e-02
Cluster-2.1  5.42  -3.58 7.51e-02
Cluster-3.0  6.31  -2.95 3.49e-02
202654 more rows ...

$comparison
[1] "1" "2"

$genes
NULL

Why there are NULL in $genes?

rna-seq no-replicate r edger
• 2,628 views • link
updated by ATpoint ATpoint • written by 1234anjalianjali1234 • 1234anjalianjali1234

This works fine for me (no errors, no warnings):

bcv <- 0.2
counts <- matrix( rnbinom(40,size=1/bcv^2,mu=10), 20,2)

Check the size of matrix:

dim(counts)
# [1] 20  2

I am guessing you have some hidden characters in the code, try to retype the commands, instead of copy-pasting from some document/website?

• 0 views • link
written by zx8754 zx8754

I have tried typing it myself, but again giving same error. as suggested by you, i have checked my counts dimention: 

> dim(counts)
    [1] 202659      2

i have tried this also but nothing worked:

counts <- data(rnbinom(40,size=1/bcv^2,mu=10), 202659,2)

i have modified my question.

• 0 views • link
written by 1234anjalianjali1234 • 1234anjalianjali1234

Why are you using data instead of matrix?

• 0 views • link
written by zx8754 zx8754

sorry it was by mistake. Even if i am using matrix instead of data, there is error:

counts <- matrix(rnbinom(40,size=1/bcv^2,mu=10), 202659,2)

Warning message:
    In matrix(rnbinom(40, size = 1/bcv^2, mu = 10), 202659, 2) :
    data length [40] is not a sub-multiple or multiple of the number of rows [202659]

Could you please explain the code, i think i am not able to set the parameters according to my dataset.

i tried:

 counts <- matrix(rnbinom(405318,size=1/bcv^2,mu=10), 202659,2)

this gives no error, but i need to understand, what this code is doing to my data and what parameter should be used like which value of mu should be taken.

Thankyou

• 0 views • link
written by 1234anjalianjali1234 • 1234anjalianjali1234

To avoid first warning, we need to supply correct dimensions, try this code, and adapt numbers for nRows, nCols as needed:

nRows <- 4
nCols <- 2
counts <- matrix(rnbinom(nRows * nCols, size = 1/bcv^2, mu = 10), nrow = nRows, ncol = nCols)

Regarding mu, read the manuals, tutorials.

• 0 views • link
written by zx8754 zx8754

Actually i need some more answers related to this question. Should i post a new question or modify this one only?

• 0 views • link
written by 1234anjalianjali1234 • 1234anjalianjali1234

1 answer

It is pointless what you do without replicates. See here (among many threads) why that is.

C: How to run the Deseq2 tool without replicates.

• 0 views • link
written by ATpoint ATpoint

In my second approach I have used my actual count data, as described in the question. Could you please look into that. I know there is no point of the analysis without the replicates, but I have no other choice.

• 0 views • link
written by 1234anjalianjali1234 • 1234anjalianjali1234

Log in to answer this question.

More posts like this