what to do when adjusted p-values are not acceptable in RNASeq?
Dear all,
I have RNASeq data of 2 normal and 5 cancer tissues.
Although the quality of reads were very good, when I run Deseq2 the adjusted p-values was not acceptable at all (only 3 genes with adjusted p- value less than 0.05).
Can you guide me how can I select differentially expressed genes in this situation?
Regards
Nazanin
• 3,408 views
•
link
written
by
nazaninhoseinkhan
1 answer
Dear JC, Panagiotis, Asaf and Macspider,
Thank you so much for your comments.
I re-run the Deseq2 after calculating read numbers using Htseq Count and the result was OK this time.
I realized that the first time I had introduced the Normal and Cancer groups wrongly.
I hope everything goes well with you
Nazanin
• 0 views
•
link
written
by
nazaninhoseinkhan
Log in to answer this question.
More posts like this
-
Problem in removing batch effect by DeSeq2
written by nazaninhoseinkhan
-
MUC5AC is not present in deseq2 results
written by nazaninhoseinkhan
-
ClusterProfiler cannot map any gene
written by nazaninhoseinkhan
-
what logFC or adj-p value cutoff should be chosen when the number of samples in one of the conditio…
written by nazaninhoseinkhan
-
Low number of healthy samples in TCGA
written by nazaninhoseinkhan
-
removing batch effects from large number of data sets
written by nazaninhoseinkhan
-
Differential co-expression analysis on TCGA transcriptomic data
written by nazaninhoseinkhan
-
Selecting Deferentially Expressed Genes in RNASeq data analysis - DESEq2 and Cuffdiff
written by Sreeraj Thamban
-
A question about RNASeq experiment design
written by nazaninhoseinkhan
-
A lot of differentially expressed genes
written by mahdijalili •
Cancer is really diverse in gene expression, could be better to analyze your cancer samples individually using the normals as "control", also you probably want to try another package as EdgeR to see if this is something particular in your dataset.
You can also try an integrated approach like metaseqR with more than one algorithms and combined p-values which may partially remedy your problem. Disclaimer: I am the author of that package.
Take a look at the correlations between the expression levels of the cancer samples (and control). You can then exclude samples that look very different from the rest or split them into two groups (according to their similarity or type).
How are you counting the reads that map on each feature? HTSeq? If yes, how many reads you have in the "NoFeature"? Are they a large fraction of your dataset?