Question: Determine Condition using LogFC values
0
gravatar for ApoorvaB
3.3 years ago by
ApoorvaB200
United States
ApoorvaB200 wrote:

Hi Everyone,

I am analyzing RNA-Seq data having 22 samples from 3 batches for differential expression of genes. My condition is test between defective and normal phenotype. For one of the samples, the condition is indeterminate. From the slides, it appears to be slightly defective but might not necessarily be so. Is there anyway I can determine what to label it as? Would it help to see changes in logFC values when i first label it as defective and then as normal?

 Thanks

fold changes rna-seq • 862 views
ADD COMMENTlink modified 3.3 years ago by dariober10.0k • written 3.3 years ago by ApoorvaB200

If you're unsure what the sample is, it is best to exclude it (not only for your interests but for others). On the other hand, if you want to look at how similar the replicates are then you can use a simple Pearson correlation values/plots to make the decision.

 

ADD REPLYlink written 3.3 years ago by arnstrm1.7k
1
gravatar for dariober
3.3 years ago by
dariober10.0k
WCIP | Glasgow | UK
dariober10.0k wrote:

Principal components analysis (PCA) is sometimes applied to expression levels from RNA-Seq data to spot outliers or otherwise unexpected sample behaviours. You could apply it to your case and see first, if samples cluster neatly by condition. Then see which cluster your undetermined sample best belongs to. Having said that, I would be careful with labelling this sample as one or the other group just on the bases of PCA. 

If you are using edgeR look at function plotMDS, I think DESeq has some similar function.

ADD COMMENTlink written 3.3 years ago by dariober10.0k
PCs on all genes will classify the unknown sample correctly, but I bet you $100 the anomalous sample lies somewhere between the two and removed in the third PC.
ADD REPLYlink written 3.3 years ago by karl.stamm3.4k
Please log in to add an answer.

Help
Access

Use of this site constitutes acceptance of our User Agreement and Privacy Policy.
Powered by Biostar version 2.3.0
Traffic: 773 users visited in the last hour