Preparing reads in TopHat

[picrinite]

Dear Friends,

When I run tophat to align RNA-Seq reads, the log file shows that some reads are discarded as below. Does anyone know why they are discarded? low quality reads? How can I setup the threshold for reads discarding?

Thank you very much.


[2015-01-25 22:24:21] Beginning TopHat run (v2.0.8b)
-----------------------------------------------
[2015-01-25 22:24:21] Checking for Bowtie
Bowtie version: 2.1.0.0
[2015-01-25 22:24:21] Checking for Samtools
Samtools version: 0.1.19.0
[2015-01-25 22:24:21] Checking for Bowtie index files
[2015-01-25 22:24:21] Checking for reference FASTA file
[2015-01-25 22:24:21] Generating SAM header for musculus
format: fastq
quality scale: phred33 (default)
[2015-01-25 22:24:44] Reading known junctions from GTF file
[2015-01-25 22:24:48] Preparing reads
left reads: min. length=101, max. length=101, 6682858 kept reads (653 discarded)
right reads: min. length=101, max. length=101, 6641398 kept reads (42113 discarded)

[Brian Bushnell]

Considering it's mainly read 2 being discarded, looks like low-quality reads. You might get better results by quality-trimming the data first, as Tophat is not very tolerant of errors; that may also reduce the number of reads discarded. Removing the last base of 101bp reads is also a good idea, IMO.

[picrinite]

Thank you.
Yes, if reads are filtered first, only 20 reads are discarded by tophat.
I have tried reading tophat manual but no information about tophat's default filtering parameters is found.