NOISeq Fold Change calculation seems off?

Question

[VBer]

I tried doing DGE analysis using NOISeq with no replicates. I got the final results as

    Control_mean    Treated_mean    M   D   prob    ranking Length  GC  Chrom   GeneStart   GeneEnd Biotype
ENSG00000000003 1726    1696    0.02529629  30  0.2973327   30.000011   2206    40.40   X   100627109   100639991   protein_coding
ENSG00000000005 1   1   0.00000000  0   0.2493902   0.000000    1205    40.78   X   100584936   100599885   protein_coding
ENSG00000000419 1674    2419    -0.53111124 745 0.7074358   -745.000189 1075    39.85   20  50934867    50958555    protein_coding
ENSG00000000457 232 251 -0.11356256 19  0.3725949   -19.000339  6308    40.14   1   169849631   169894267   protein_coding
ENSG00000000460 420 452 -0.10593344 32  0.3947228   -32.000175  3849    39.22   1   169662007   169854080   protein_coding
ENSG00000000938 2   1   1.00000000  1   0.4415171   1.414214    2637    52.92   1   27612064    27635185    protein_coding

The Ensembl IDs are the row names. The Mean values are the mean counts across replicates (since there are no replicates they remain the same as the original counts). If you look at the M value, which according to the manual is the log2 ratio between the two conditions, it is calculated as log2(control/treated) by NOISeq as opposed to log2(treated/control). I don't know why this is happening because my counts and factor data frames are proper.

The counts are:

Treated_S2  Control_S1
ENSG00000000003 1696    1726
ENSG00000000005 1   1
ENSG00000000419 2419    1674
ENSG00000000457 251 232
ENSG00000000460 452 420
ENSG00000000938 1   2

The factors are:

group
Treated
Control

The command I used to create the NOISeq data object is:

myresults <- noiseq(mydata, factor = "group", k = NULL, norm = "n", pnr = 0.2, nss = 5, v = 0.02, lc = 1, replicates = "no", conditions = c("Treated","Control"))

Somehow NOISEq swaps the control and treated columns and calculates the fold change accordingly. Can someone tell me what's going on?

Answer 1

[Hyper_Odin]

While calculating it considers the alphabetical order of the group you have defined, so, in your case it will calculate Control vs Treated, Change the names and you are good to go!!

[VBer]

Ooh, yes, let me check and let you know how it goes!

[Hyper_Odin]

Also be careful with the sample names. Double check if the samples names correspond to the group you have assigned.