Hi everyone
By analyzing my data I've come across a strange circumstantial inconsistency with FPKM values in the cuffdiff summary file vs. the genes.read.group.tracking file. I'll try to provide as much background and detail as I can. Sorry for it being long winded.
Background
HiSEq 50SR Illumina adapters, 6 mulitiplexed samples per lane.
Analysis pipeline - Align with Tophat, use cufflinks/cuffdiff for differential expression.
I am comparing the effects of certain gene knockouts (KO) on the transcriptome with wild type.
What I want to do
I want to filter out genes that have low read numbers from my list of ~10,000 genes in the cuffdiff output file.
I think I should be able to calculate these values, but aligning the transcript sizes with the output files has been annoying because there is a small percentage of hits on the cuffdiff output that provide multiple gene names, which throws the alignment off in the spreadsheet from the gtf file, which requires me to manually go through the data and align the rows (not fun)
During the process of, quality control, directly comparing my biological replicates to each other (e.g. WT1 vs WT2) using cuffdiff I found that I could align raw reads from the genes.read.group.tracking file to the corresponding cuffdiff output using the FPKM values (the FPKMs are identical between the two files). This is what I am doing with the data
This worked wonderfully for pairwise comparisons between my biological replicates (all 6 of them). e.g. WTa vs WTb or KO1a vs KO1b.
However, when I tried the same technique on my cuffdiff outputs (comparing WT to a knockout, each with two biological replicates -- WTa,WTb vs KOa,KOb) the FPKMS from the genes.read.group.tracking file do not match up at all with the FPKMs in the cuffdiff summary (neither in the mean FPKM columns or the individual FPKM columns). In other words I could not align the raw counts to the genes. The total number (quantity) of unique 'genes/FPKMs' in the genes.read.group.tracking file matches the total number in the cuffdiff summary, but none of the FPKMs numerically match.
***All the analysis pipelines, options, tophat files, etc. are identical in both types of experiments (comparing WTa vs WTb or or comparing KO1a vs KO1b or comparing WTa,WTb vs KO1a,KO1b)***
This doesn't make any sense to me as I was under the impression that the tracking files all need to correspond to their respective cuffdiff summary.
Does anyone know why there is this inconsistency?
Can anyone suggest a way to fix it or provide an alternative to aligning my cuffdiff output with raw reads?
thanks in advance
By analyzing my data I've come across a strange circumstantial inconsistency with FPKM values in the cuffdiff summary file vs. the genes.read.group.tracking file. I'll try to provide as much background and detail as I can. Sorry for it being long winded.
Background
HiSEq 50SR Illumina adapters, 6 mulitiplexed samples per lane.
Analysis pipeline - Align with Tophat, use cufflinks/cuffdiff for differential expression.
I am comparing the effects of certain gene knockouts (KO) on the transcriptome with wild type.
What I want to do
I want to filter out genes that have low read numbers from my list of ~10,000 genes in the cuffdiff output file.
I think I should be able to calculate these values, but aligning the transcript sizes with the output files has been annoying because there is a small percentage of hits on the cuffdiff output that provide multiple gene names, which throws the alignment off in the spreadsheet from the gtf file, which requires me to manually go through the data and align the rows (not fun)
During the process of, quality control, directly comparing my biological replicates to each other (e.g. WT1 vs WT2) using cuffdiff I found that I could align raw reads from the genes.read.group.tracking file to the corresponding cuffdiff output using the FPKM values (the FPKMs are identical between the two files). This is what I am doing with the data
This worked wonderfully for pairwise comparisons between my biological replicates (all 6 of them). e.g. WTa vs WTb or KO1a vs KO1b.
However, when I tried the same technique on my cuffdiff outputs (comparing WT to a knockout, each with two biological replicates -- WTa,WTb vs KOa,KOb) the FPKMS from the genes.read.group.tracking file do not match up at all with the FPKMs in the cuffdiff summary (neither in the mean FPKM columns or the individual FPKM columns). In other words I could not align the raw counts to the genes. The total number (quantity) of unique 'genes/FPKMs' in the genes.read.group.tracking file matches the total number in the cuffdiff summary, but none of the FPKMs numerically match.
***All the analysis pipelines, options, tophat files, etc. are identical in both types of experiments (comparing WTa vs WTb or or comparing KO1a vs KO1b or comparing WTa,WTb vs KO1a,KO1b)***
This doesn't make any sense to me as I was under the impression that the tracking files all need to correspond to their respective cuffdiff summary.
Does anyone know why there is this inconsistency?
Can anyone suggest a way to fix it or provide an alternative to aligning my cuffdiff output with raw reads?
thanks in advance