Loss of relevant genes in NGS...
Dear SEQanswers community,
I have a quite fundamental concerns regarding using NGS to explore differentially expressed genes.
I am wondering whether some of the important (like clinically important) genes with relatively low expression might actually be undetectable in NGS analysis.
Take my case as example, I have 3 cell groups, which i want to apply whole transcriptome sequencing to distinguish them. Before I performed NGS, I used qPCR to target three genes which were known to have different expression level in these three cell groups. The qPCR showed a very clear result that these 3 cell groups do have different expression of these 3 genes.
Now I have done the NGS and want to see whether the expression of these 3 genes in these 3 cell groups align with my qPCR results.
1 out of the 3 genes had a normalized count of 0 in all cell groups, so I cannot even compare. I was able to compare the other 2, and the results seem to point to the same direction as qPCR does, but the standard deviation is quite large.
I am wondering whether during the normalization of featureCounts, the count of these three genes were being normalized to 0 or close to 0, because other house-keeping genes are far more abundant than the three genes I am looking at. Which means, with NGS, I actually loss the information of some important genes due to there relatively low expression, despite clinical or pathological impacts from them are high.
To give more information for assessment, I performed RNA STAR with trimmed fastq reads, and used featureCounts to count the gene expression. Count normalization was done with DeSeq2. I simply extracted the rLog(normalized counts) of these genes to excel and used simple excel functions to do statistics.
The same thing has been done with edgeR, and the normalized count results were used. Similar results...
The gene which had 0 normalized count was TNF-alpha.
The other two were IL6 and TGFb1.
All are quite well-known inflammatory cytokines.
Thanks in advance for your help.
Susan
I have a quite fundamental concerns regarding using NGS to explore differentially expressed genes.
I am wondering whether some of the important (like clinically important) genes with relatively low expression might actually be undetectable in NGS analysis.
Take my case as example, I have 3 cell groups, which i want to apply whole transcriptome sequencing to distinguish them. Before I performed NGS, I used qPCR to target three genes which were known to have different expression level in these three cell groups. The qPCR showed a very clear result that these 3 cell groups do have different expression of these 3 genes.
Now I have done the NGS and want to see whether the expression of these 3 genes in these 3 cell groups align with my qPCR results.
1 out of the 3 genes had a normalized count of 0 in all cell groups, so I cannot even compare. I was able to compare the other 2, and the results seem to point to the same direction as qPCR does, but the standard deviation is quite large.
I am wondering whether during the normalization of featureCounts, the count of these three genes were being normalized to 0 or close to 0, because other house-keeping genes are far more abundant than the three genes I am looking at. Which means, with NGS, I actually loss the information of some important genes due to there relatively low expression, despite clinical or pathological impacts from them are high.
To give more information for assessment, I performed RNA STAR with trimmed fastq reads, and used featureCounts to count the gene expression. Count normalization was done with DeSeq2. I simply extracted the rLog(normalized counts) of these genes to excel and used simple excel functions to do statistics.
The same thing has been done with edgeR, and the normalized count results were used. Similar results...
The gene which had 0 normalized count was TNF-alpha.
The other two were IL6 and TGFb1.
All are quite well-known inflammatory cytokines.
Thanks in advance for your help.
Susan