Tweet
Hi Everyone,
When I align my reads with tophat against the GRCm38.fa and then use various programs (including Cufflinks) to match them to known features using the Ensembl gtf file, many of my reads (~50%) are unassigned to known features. Therefore, someone suggested I use the transcripts.gtf file produced by Cufflinks to build a new/modified annotation gtf which finds groups of reads that are aligned to the genome but in an area that is not associated with a known transcript.
Therefore, I've been trying to build a new/modified annotation gtf file for Mus musculus specific for my cells of interests. I've tried to do this a few different ways using Cufflinks.
1. I've pooled all my reads and used the transcripts.gtf file produced by Cufflinks using the -g option
Pros: I boost the rate of alignment to features to over 80%
Cons: these are huge data files that take forever to process, don't contain all the original Ensembl.gtf transcripts
2. I've used cuffmerge on all the individual transcripts.gtf files produced by my individual samples
Pros: I boost the rate of alignment to features to over 60%
Cons: cuffmerge strips a lot of the data contained in the initial Ensembl.gtf file as well as only contains transcripts expressed by my cells.
3. I've run cufflinks with no -g option and then used cuffmerge to try and link the Ensembl.gtf file with the transcripts.gtf file
Pros: I boost the rate of alignment to features to over 60%
Cons: Because the output files are so different from the original input, it's hard to compare but there are 400 000 more lines in my original Ensembl.gtf than in the cuffmerged files. Where did all that information go?
My two biggest concerns is that one, my new gtf files don't contain all the transcripts in the original Ensembl file, and two Cufflinks removes features and relabels a lot of the attributes like gene_id.
Is there a better way to do this? Should I just concatenate the files together? I'm not sure what the best approach is.
Thanks all for your help.
When I align my reads with tophat against the GRCm38.fa and then use various programs (including Cufflinks) to match them to known features using the Ensembl gtf file, many of my reads (~50%) are unassigned to known features. Therefore, someone suggested I use the transcripts.gtf file produced by Cufflinks to build a new/modified annotation gtf which finds groups of reads that are aligned to the genome but in an area that is not associated with a known transcript.
Therefore, I've been trying to build a new/modified annotation gtf file for Mus musculus specific for my cells of interests. I've tried to do this a few different ways using Cufflinks.
1. I've pooled all my reads and used the transcripts.gtf file produced by Cufflinks using the -g option
Pros: I boost the rate of alignment to features to over 80%
Cons: these are huge data files that take forever to process, don't contain all the original Ensembl.gtf transcripts
2. I've used cuffmerge on all the individual transcripts.gtf files produced by my individual samples
Pros: I boost the rate of alignment to features to over 60%
Cons: cuffmerge strips a lot of the data contained in the initial Ensembl.gtf file as well as only contains transcripts expressed by my cells.
3. I've run cufflinks with no -g option and then used cuffmerge to try and link the Ensembl.gtf file with the transcripts.gtf file
Pros: I boost the rate of alignment to features to over 60%
Cons: Because the output files are so different from the original input, it's hard to compare but there are 400 000 more lines in my original Ensembl.gtf than in the cuffmerged files. Where did all that information go?
My two biggest concerns is that one, my new gtf files don't contain all the transcripts in the original Ensembl file, and two Cufflinks removes features and relabels a lot of the attributes like gene_id.
Is there a better way to do this? Should I just concatenate the files together? I'm not sure what the best approach is.
Thanks all for your help.