gpertea
commented
7 years ago Thank you for providing the .sam files but it seems that I am not able to reproduce the issues with just the provided data. I simply ran stringtie -o s163.gtf Sample163_hisat.sam and the output shows large FPKM and coverage values:
# stringtie -o s163.gtf Sample163_hisat.sam
# StringTie version 1.3.3b
chr3 StringTie transcript 181712403 181715282 1000 + . gene_id "STRG.1"; transcript_id "STRG.1.1"; cov "242.097504"; FPKM "382242.812500"; TPM "497675.312500";
chr3 StringTie exon 181712403 181714359 1000 + . gene_id "STRG.1"; transcript_id "STRG.1.1"; exon_number "1"; cov "254.133179";
chr3 StringTie exon 181715185 181715282 1000 + . gene_id "STRG.1"; transcript_id "STRG.1.1"; exon_number "2"; cov "1.752461";
chr3 StringTie transcript 181711860 181712281 1000 . . gene_id "STRG.2"; transcript_id "STRG.2.1"; cov "236.135071"; FPKM "372828.843750"; TPM "485418.468750";
chr3 StringTie exon 181711860 181712281 1000 . . gene_id "STRG.2"; transcript_id "STRG.2.1"; exon_number "1"; cov "236.135071";So the problem might be context-related. What was the stringtie command line you used? Did you run with the "-e" option (it seems to) ? In that case I would also need the reference transcripts you used with the -G option.
It would be helpful if you could follow the instructions provided here to extract the read alignments and the reference transcripts for a region of interest and also provide the exact command line you used. I hope the version of stringtie you used is also the latest available (v1.3.3b).
MassaBob
Yunqing-Yu
SimonJYoung
zhanxiaoyu
freedog8
OwenDonohoe
I used stringtie to analyse the output of my hisat alignment in >30 human samples in parallel and detected single samples with FPKM values of 0 for several genes although the regions are well covered by reads.
This seems to resemble https://github.com/gpertea/stringtie/issues/102. Using cufflinks I get FPKM values as expected.
For troubleshooting I picked two of the samples showing 0PKM in my favourite gene SOX2 and two samples showing high FPKMs and repeated the analysis using only reads mapping to the gene. I could recapitulate the error. I changed my pipeline using STAR aligner for mapping followed by stringtie analysis and this led to FPKM values >0 in three of the four samples, i.e. recovering one sample. Applying Cufflinks instead of stringtie showed high FPKM values for all 4 samples.
Attached you will find sam files created with hisat and STAR covering the SOX2 region for one "good" sample (Sample 81) and the sample that could be rescued using STAR aligner (sample 163).
SAM_ARCHIVE.zip
SOX2 is well covered in both samples:
Furthermore, I added gtf output for the two samples of all four different analysis combinations (hisat vs STAR + Stringtie vs cufflinks) with the SOX2 gene extracted. SOX2_gtfs.zip
I appreciate any help or any hint what might have happened here and how to deal with these (recurring) issues in the future.