Documentation query

[AMChalkie]

Hi SuperTranscribers,

Nice work, looking at using this for visualisation of our data.

Silly question: in the bash_commands.txt you reference a genome GTF file:
GenomeAnno=ST_20120709_trinity_corset/SuperDuper.gtf

Then subsequently don't use it in the STAR genome generation (--sjdbGTFfile ). Any reason you don't give STAR a gtf to guide it?

Also, there are points in the bash script where it is unclear which bits are chicken and which are human and extra clarification would be helpful.

Thanks
/Alistair

[Quarkins]

Hi Alistair,

The bash commands.txt was just kind of a dump from some of scripts we had written for the analysis so apologies if they aren't too clear. They should all be relevant to Human (the chicken stuff is in a seperate folder). But some of it was simulation and some was real data.

As for the SuperDuper.gtf you are right, you should pass the GTF file to STAR to help guide it.

Are you working trying to align reads to a SuperTranscript made from DeNovo assembled reads or form the Human super reference? We are happy to clarify anything in the commands, just let me know which parts don't make sense.

Cheers
/Anthony

[AMChalkie]

Hi Anthony,

My use for this are to visualise human/mouse alternative splicing events, so thats the area I am interested in. Particularly given the human_super_reference.fasta for human, the steps for running the Trapnell dataset as a clearly defined section of code would be useful. SuperDuper.fasta is referenced for chicken, and human_super_reference.fasta for human. Are these equivalent files?

[Quarkins]

Ok great,
So the human_super_reference.fasta is simply the SuperTranscripts made by concatenating all known exon sequence for each gene, so it is not built using the Lace software but by the annotation. So i think for your purposes you just want to map reads using purely this as your reference.

The SuperDuper.fasta (in the Chicken directory) are the supertranscripts constructed by Lace using a mixture of annotated transcripts, genome guided assembly and de novo assembly. So this you most definitely don't want to use, unless you are studying chickens.

As for the Trapnell dataset, in order to align the reads to the human super transcriptome, you would use a standard STAR 2 pass alignment aligning to human_super_reference.fasta.

`##########################################

Running STAR

##########################################

#Make Index for huamn_super_reference.fasta
STAR --runMode genomeGenerate --runThreadN 16 --genomeDir STARIndex --genomeFastaFiles human_super_reference.fasta --limitGenomeGenerateRAM 75070378709

#Get the files
firstpass=1pass
secondpass=pass
GenomeDir=STARIndex
outdir=
GenomeFasta=human_super_reference.fasta
GenomeAnno=ST_20120709_trinity_corset/SuperDuper.gtf

#First pass mapping
cd /group/bioi1/shared/public_data/cuffdiff2_data/

STAR --genomeDir $GenomeDir
--readFilesIn SRR493366_1.trimmed.fastq,SRR493367_1.trimmed.fastq,SRR493368_1.trimmed.fastq,SRR493369_1.trimmed.fastq,SRR493370_1.trimmed.fastq,SRR493371_1.trimmed.fastq SRR493366_2.trimmed.fastq,SRR493367_2.trimmed.fastq,SRR493368_2.trimmed.fastq,SRR493369_2.trimmed.fastq,SRR493370_2.trimmed.fastq,SRR493371_2.trimmed.fastq
--outFileNamePrefix $firstpass/SRR
--outSAMtype BAM Unsorted --runThreadN 24;

#Generate genome with junctions from 1st pass
STAR --genomeDir $secondpass --runMode genomeGenerate --genomeFastaFiles $GenomeFasta --sjdbFileChrStartEnd $firstpass/SRRSJ.out.tab --sjdbOverhang 100 --runThreadN 20 --limitGenomeGenerateRAM 75070378709;

#Run 2nd pass mapping to the new genome defined from sjdb
ls *.fastq | rev | cut -c 17- | rev | uniq > myfiles
for i in cat myfiles; do
echo "Performing alignment on file $i"
STAR --genomeDir $secondpass
--readFilesIn "$i"_1.trimmed.fastq "$i"_2.trimmed.fastq
--outFileNamePrefix $outdir/"$i"
--outSAMtype BAM Unsorted --runThreadN 20 --outSJfilterOverhangMin 12 12 12 12;
done
'

I just copied the commands from the bash_commands.txt and modified the name to map to the human_super_reference.fasta (rather than Superduper.fasta), then you should be good to go. Note, the fastq files from the Trapnell dataset were trimmed using Trimmomatic in the standard way, you don't have to do that but it maybe cleaner.

Then visualise the BAM files either in IGV (using human_super_reference.fasta as your reference genome).

Hope that makes sense,
/Anthony