fastq_screen vs. tophat2

[frymor]

Hi all,

I was wondering about why I get a relatively low percentage of reads mapped to my genome.
We are running an experiment of several fruit fly samples. we have several time-points and knock-outs.
the fastq files are around 100M reads, but unfortunately I am able to map between 65-75% of them.

To check for contamination I have tried the tools fastQ_screen. Apparently my data set is good (s. image).

The quality of the data is very good and the according to my fastqc results, I don't have any noteworthy duplications in the data.
What I don't understand is how bowtie2 in the fastq_screen can assign >98% of the reads to the fly genome, but in tophat2 run I get only ~65%.

These are the command I have used to run both tophat2 and fastq_screen:

tophat -p 10 -G genes.gtf -o A ~/genomes/Drosophila_melanogaster/Ensembl/BDGP6.80/bowtie2index/genome 28023_TGACCA_C7F7GANXX_5_20150619B_20150619.fastq

and

./fastq_screen --force --nohits --subset 0 --outdir B --aligner bowtie2 --threads 14 28025_CGATGT_C7F7GANXX_8_20150619B_20150619.fastq

Does bowtie2 uses different parameters in fastq_screen than in the tophat2 run?
Is there a way to increase the mapping results?

thanks for any ideas or hints,
Assa

[Brian Bushnell]

Tophat is not very sensitive by default. You could have low-quality data, or adapter contamination; in either case, those reads would generally not map. I suggest you try adapter-trimming and quality-trimming, to see if that improves things.

You can also dramatically increase the sensitivity by using BBMap instead of TopHat. For RNA-seq in animals, I suggest adding the flag "maxindel=200k"; otherwise, the defaults are fine:
bbmap.sh in=reads.fq out=mapped.sam ref=fly.fa maxindel=200k