Entering edit mode
5 months ago
junli1988
•
0
Hi,
I have an Excel sheet of RNAseq results. The raw data was analyzed by a company and was not available to me. I have three sets of data on this sheet besides the gene names and p-values: Raw counts, read counts (normalized, I think?) and FPKM.
I want to pick out a couple of significant DE genes for a bar graph to empathize their expression levels. Which data should I use for this purpose? The FPKM or a log10 variation of the normalized counts? I'm not a biostatistician by any means, just a neuroscientist. Plus, this data is second/third-hand stuff (from company to collaborating PI to my PI). Can someone help me parse through this stuff?
This is the stuff of nightmares. Please save the datasets individually in plain text, ensure gene names were not massacred by Excel and use the plain text files.
LOL...It's a CSV file, and that's all I have in the folder. I agree that it is indeed the stuff of nightmares. But what can I do? It's not too bad since the file only contains genes with a padjust < 0.05 (about 50 or so genes). I checked the gene names manually so we could get the Excel butchering out of the way first.
Can you do me a favor and tell me which count to use? What's the meaning of FPKM? Does it have any real use? What's your answer to my question? Log10 of the normalized read counts? I know we use normalized read counts in transcriptional data from NanoString and other mRNA assays. But this FPKM thing in RNAseq really throws me off. I read through a couple of posts on this site, but I'm still unsure what to use.
People use FPKM and TPM for analyses but it's best to use raw counts. That being said, if the file already only contains genes with padj < 0.05, why do you need to do anything further to identify significant DE genes? A list of genes filtered by log2FC and p.adj is the end product of DE calculations.