How can i extract FPKM from my GTF file which is created by stringtie?

Question

[Zeason]

I just do some practice on ballgown and stringtie, and I got some GTF file and ballgown`s file.

However, I just find I can't use R or something else to deliver a excel file which contain FPKM.

I think the excel file I want maybe looks like this:

gene_id      FPKM
    A        124
    B        541   
    C        122

Please help me

Thanks a lot :)

[vkkodali_ncbi]

Can you paste a few sample lines from the GTF file you are working with?

[Zeason]

this is the top ten lines:

1   StringTie   transcript  337772  338047  1000    +   .   gene_id "Zm00001d027250"; transcript_id "Zm00001d027250_T001"; cov "66.076088"; FPKM "8.407302"; TPM "14.141658";
1   StringTie   exon    337772  338047  1000    +   .   gene_id "Zm00001d027250"; transcript_id "Zm00001d027250_T001"; exon_number "1"; cov "66.076088";
1   StringTie   transcript  426764  432130  1000    +   .   gene_id "Zm00001d027254"; transcript_id "Zm00001d027254_T001"; cov "2.043083"; FPKM "0.259955"; TPM "0.437262";
1   StringTie   exon    426764  426798  1000    +   .   gene_id "Zm00001d027254"; transcript_id "Zm00001d027254_T001"; exon_number "1"; cov "0.000000";
1   StringTie   exon    426869  426970  1000    +   .   gene_id "Zm00001d027254"; transcript_id "Zm00001d027254_T001"

i think maybe a python script can extract the fpkm , but i dont know how to edit a complex python script.so i want to find some software to do this work.thanks a lot

Answer 1

[vkkodali_ncbi]

You can use standard unix commands for this as follows:

$ cat stringtie.txt
1  StringTie  transcript  337772  338047  1000  +  .  gene_id "Zm00001d027250"; transcript_id "Zm00001d027250_T001"; cov "66.076088"; FPKM "8.407302"; TPM "14.141658";
1  StringTie  exon        337772  338047  1000  +  .  gene_id "Zm00001d027250"; transcript_id "Zm00001d027250_T001"; exon_number "1"; cov "66.076088";
1  StringTie  transcript  426764  432130  1000  +  .  gene_id "Zm00001d027254"; transcript_id "Zm00001d027254_T001"; cov "2.043083"; FPKM "0.259955"; TPM "0.437262";
1  StringTie  exon        426764  426798  1000  +  .  gene_id "Zm00001d027254"; transcript_id "Zm00001d027254_T001"; exon_number "1"; cov "0.000000";
1  StringTie  exon        426869  426970  1000  +  .  gene_id "Zm00001d027254"; transcript_id "Zm00001d027254_T001"
$ grep 'FPKM' temp.txt | cut -f9 | sed -r 's/gene_id "([^"]*).*FPKM "([^"]*).*/\1\t\2/g' | sed '1i#gene_id\tFPKM'
#gene_id        FPKM
Zm00001d027250  8.407302
Zm00001d027254  0.259955

[Zeason]

thank you very much , i will try

[waqaskhokhar999]

How can we get transcript level TPM values instead of gene level TPM values, I have tried to replace gene_id with transcript_id but it didn't work for me?

[vkkodali_ncbi]

Assuming that your GTF file is same as above, you can do the following:

$ grep 'TPM' temp.txt | cut -f9 | sed -r 's/.*transcript_id "([^"]*).*TPM "([^"]*).*/\1\t\2/g' | sed '1i#transcript_id\tTPM'

[waqaskhokhar999]

Many thanks for your response, it work fine most of the lines till the pattern sustains like:

gene_id "MSTRG.26629"; transcript_id "AT5G53360.2"; cov "14.228090"; FPKM "5.268032"; TPM "8.616198";

But generates error when ref_gene_name in 9th column contains TPM letters in gene name (ATPMEPCRF)

gene_id "MSTRG.26631"; transcript_id "AT5G53370.1"; ref_gene_name "ATPMEPCRF"; cov "279.969208"; FPKM "103.660202"; TPM "169.542801";

Can you please check this issue?

[EagleEye]

Simple solution will be,

grep -w "TPM"

OR

grep " TPM "

OR

grep "; TPM "

Answer 2

[EagleEye]

Form StringTie output you can use 'sample1_gene_abund.tab' file to extract these information,

FPKM:

cat {sample1}_gene_abund.tab | cut -d$'\t' -f1,8 | sed "s/FPKM/sample1_FPKM/" | sed 's/Gene ID/gene_id/' > {sample1}_gene_abund.fpkm.txt

TPM:

cat {sample1}_gene_abund.tab | cut -d$'\t' -f1,9 | sed "s/TPM/sample1_TPM/" | sed 's/Gene ID/gene_id/' > {sample1}_gene_abund.tpm.txt

[Zeason]

thank you firstly ,i will try it i got another question here , i always think the gene FPKM is a sums of its all transcript FPKM , is that right ? thanks a lot

[EagleEye]

It is not always the case. There are also cases like this where it depends on the quantification approach,

Image publication ref

[Zeason]

really really thank you very much , i got it.