Dear all,
It would be much more useful if anyone could help me with this unusual trend. I have had three types of samples (Control, a transcription factor -overexpressed and a transcription factor -silenced) with 4 replicates each. My objective was to identify the genes in a particular pathway in soybean that are regulated by this transcription factor. Genes with more than 2-fold differentially expressed were chosen. RNA-seq analyses were done by EdgeR and CLC bio-genomic workbench (similar results in both ways). However, when I was validating some of the differentially expressed genes by qPCR, I found that three of the genes showed totally opposite trend of RNA-seq data. The transcription factor and one of its target genes were correlated perfectly by QPCR but one of the genes that is 40-fold downregulated in overexpressed lines by RNA-seq, is ~40-fold upregulated by QPCR. Is this really normal? Little bit change in fold change would be normal, but switching between upregulation and downregulation in two methods seems unbelievable! (especially 40-fold!!)
Typically, RNASeq data is generated by summarizing the reads over the mRNA of each gene. In qPCR you typically design a primer pair over a smallish (100 bp) piece of the entire mRNA. These different approaches might be causing the phenomenon you see.
It would be good to check the aligned reads of your RNASeq experiment (the .bam files) in IGV to see whether the sequence you investigated with qPCR is shows the same trend as the overall reads on the gene. Otherwise, the differences might be explained by differential splicing. In Ensembl, you might have a look at the genes that are giving you trouble to see whether they have splice variants that might cause you trouble.
You might also want to look into the DEXSeq bioconductor package, that is made for looking at differential splicing.
https://bioconductor.org/packages/release/bioc/html/DEXSeq.html
How about the TFBS? Have you performed something liken an EMSA to confirm binding of the TF to the respective promoters/putative binding sites? This would be helpful to decide which result one should better trust.
There can a few things go wrong in qPCR, and many things can go wrong in RNAseq (starting with the library prep, through all the heuristics used to rate the quality of the reads, up to the alignment on the reference genome). It won't be possible to find the reason for this discrepancy by writing posts in RG. Better ask some experts at your place, go through the experiments and through the (raw!) data and all processing steps to see where there might be problems.
Another option you have is to quantify the protein (if possible), e.g. in western blots. Even a semi-quantitative histological analysis should be sufficient (a 40-fold difference should easily be visible). It bears the risk that the protein is not regulated (despite the fact that the RNA is) - but if you find it regulated, this is a much more valuable evidence than qPCR and RNAseq together.
Typically, RNASeq data is generated by summarizing the reads over the mRNA of each gene. In qPCR you typically design a primer pair over a smallish (100 bp) piece of the entire mRNA. These different approaches might be causing the phenomenon you see.
It would be good to check the aligned reads of your RNASeq experiment (the .bam files) in IGV to see whether the sequence you investigated with qPCR is shows the same trend as the overall reads on the gene. Otherwise, the differences might be explained by differential splicing. In Ensembl, you might have a look at the genes that are giving you trouble to see whether they have splice variants that might cause you trouble.
You might also want to look into the DEXSeq bioconductor package, that is made for looking at differential splicing.
https://bioconductor.org/packages/release/bioc/html/DEXSeq.html
In qPCR you may need to use more than one reference gene, since this could be the cause of the problem.
In RNA-Seq, the problem could be caused by what Martijn suggest. You may want to check for overall changes in expression, trying to find genes that may have high increase in expression that will affect the relative counts of your genes. Remember that in RNA-Seq, sequencing is at random, and if there are more transcripts from other genes, your sequences will be sampled less often.
- Badges
- Science topic