Hi,
I see problems with the fpkm values calculated by cufflinks program. I dont see correlation between counts and fpkms for some genes.
To be more clear I explain my scenario
I have RNA-SEQ data for two conditions (let say condA and condB) with 3 replicates each
for a particular gene : say gene Tagap
Cond A:
Type Rep1 Rep2 Rep3 mean_of_replicates
Count 169 171 124 154
fpkms 4.44 3.8 3.2 3.8
Cond B:
Type Rep1 Rep2 Rep3 mean_of_replicates
Count 17 38 21 25
fpkms 5.0 4.7 4.3 4.6
From above :
CondA (counts) > CondB (counts)
CondA (fpkms) < CondB (fpkms)
I tried to understand the fpkm calculations from cufflinks supplementary methods, but I couldnt completely get their calculations. After googling abit, I thought I can use this simple formulae to do normalization
Nc = (10^9*Counts)/(Transcript_length*Total no. of uniquely mapped reads)
with this formulae, I made use of counts to make normalizations and got below values
Cond A:
Type Rep1 Rep2 Rep3 mean_of_replicates
Nc 1.72 1.88 1.26 1.62
Cond B:
Type Rep1 Rep2 Rep3 mean_of_replicates
Nc 0.18 0.4 0.26 0.28
CondA (Nc) > CondB (Nc)
Nc values correlates with the counts.
Can I use this approach (Nc calculations) to get the normalized values instead of FPKMs calculated from cufflinks?
Why the direction of fold change is completely reversed by cufflinks fpkms?
I see problems with the fpkm values calculated by cufflinks program. I dont see correlation between counts and fpkms for some genes.
To be more clear I explain my scenario
I have RNA-SEQ data for two conditions (let say condA and condB) with 3 replicates each
for a particular gene : say gene Tagap
Cond A:
Type Rep1 Rep2 Rep3 mean_of_replicates
Count 169 171 124 154
fpkms 4.44 3.8 3.2 3.8
Cond B:
Type Rep1 Rep2 Rep3 mean_of_replicates
Count 17 38 21 25
fpkms 5.0 4.7 4.3 4.6
From above :
CondA (counts) > CondB (counts)
CondA (fpkms) < CondB (fpkms)
I tried to understand the fpkm calculations from cufflinks supplementary methods, but I couldnt completely get their calculations. After googling abit, I thought I can use this simple formulae to do normalization
Nc = (10^9*Counts)/(Transcript_length*Total no. of uniquely mapped reads)
with this formulae, I made use of counts to make normalizations and got below values
Cond A:
Type Rep1 Rep2 Rep3 mean_of_replicates
Nc 1.72 1.88 1.26 1.62
Cond B:
Type Rep1 Rep2 Rep3 mean_of_replicates
Nc 0.18 0.4 0.26 0.28
CondA (Nc) > CondB (Nc)
Nc values correlates with the counts.
Can I use this approach (Nc calculations) to get the normalized values instead of FPKMs calculated from cufflinks?
Why the direction of fold change is completely reversed by cufflinks fpkms?
Comment