expression profiles between DESeq2 and Mfuzz
Entering edit mode
Assa Yeroslaviz ★ 1.5k
Last seen 12 weeks ago


we're working on a data set from D. melanogaster and examining a series of TP of different knock-outs. we have several RNA-Seq data sets with triplicates. I was analyzing both the differences in each pair of time-points using DESeq2 as well as the time-series analysis using Mfuzz.

I have tried to compare the expression values (or more precise profiles) created by Mfuzz and DESeq2 and got confusing results. Below is an example for one gene, its raw values as well as the calculated values from Mfuzz and normalised values from DESeq2

TP gene1 raw gene1.mfuzz mean / standadised gene1 deseq2. norm gene1 deseq2 log2(mean)
TP0.1 8141   5800.295958  
TP0.2 3229 5928 6495.528088 12.59127865
TP0.3 6416 -0.478593661 6217.048568  
TP1.1 3806   3012.129665  
TP1.2 8203 5517 2201.374318 11.18855547
TP1.3 4544 -0.535847289 1788.334031  
TP2.1 5762   9765.202527  
TP2.2 23550 16900 6191.957196 12.8355461
TP2.3 21389 1.049841384 5971.174465  
TP3.1 28123 23003 13424.23626 13.45799774
TP3.2 17883 1.900008999 9081.384553  
TP4.1 1073   4857.789599  
TP4.2 9001 6998 6889.992045 12.71665802
TP4.3 10922 -0.329539205 8445.958513  
TP5.1 12129 10613 20796.96629 14.02112321
TP5.2 9097 0.174041974 12454.33575  
TP6.1 3988 4095 8288.285683 13.53372723
TP6.2 4203 -0.733936483 15430.24845  
TP7.1 1542 1855 4384.120254 12.12940242
TP7.2 2168 -1.045975719 4576.621367  


Plotting these values gives a completely different expression behavior of the same gene across all time-points (see plots attached).


I would like to know, if these two different expression profiles can be solely explained by the difference between the DESeq2 normalisation and the Mfuzz standardisation methods. If not, is there another explanation for this two plots?

thanks in advance 


deseq2 mfuzz differential expression timecourse • 1.3k views
Entering edit mode
Last seen 7 hours ago
United States

DESeq2's normalized counts are pretty straightforward: these are counts / size factors, where the size factors account for differences in sequencing depth across samples. The normalized counts are then scaled count values, with the lowest and highest sequencing depth samples scaled up- and down- respectively to be similar scale to the middle samples.

I don't know anything about the normalization of Mfuzz. You can email the authors of that method for more information if it is not explained in their documentation or paper.

(Remember, the normalized counts in DESeq2 are just provided for visualization, they aren't actually used in the DESeq2 model which uses all samples on the original counts scale and keeps track of size factors on the right-hand side of the equations as shown in the DESeq2 paper.)


Login before adding your answer.

Traffic: 389 users visited in the last hour
Help About
Access RSS

Use of this site constitutes acceptance of our User Agreement and Privacy Policy.

Powered by the version 2.3.6