Replicating bowtie2 alignment parameters with Rsubread
1
0
Entering edit mode
@dktanwar1991-7583
Last seen 22 months ago
United States

I would like to use Rsubread instead of bowtie2. Following are my parameters for bowtie2:

-X 2000 --end-to-end --no-mixed --no-discordant --score-min L,-0.4,-0.4

How can I replicate them with Rsubread?

-X --> maxFragLength

--end-to-end --> ?

--no-mixed --> ?

--no-discordant --> ?

--score-min L,-0.4,-0.4 --> ?

1
Entering edit mode
Wei Shi ★ 3.3k
@wei-shi-2183
Last seen 1 day ago
Australia/Melbourne/Olivia Newton-John …

The align and subjunc functions in Rsubread always soft-clip reads if they find end bases in a read cannot be mapped, therefore they do not support the bowtie2 parameter '--end-to-end'. Rsubread aims to report confident mapping results for each base in each read. If end bases in a read can not be confidently mapped, Rsubread chooses to soft-clip them instead of reporting locations that are likely to be wrong.

Regarding '--no-mixed', this is the default behavior of align and subjunc. Rsubread aligners always try to map the two reads from a pair individually if they cannot map them as a pair.

Rsubread aligners do not support '--no-discordant'. align and subjunc will look for discordant alignments for a read pair if they cannot find a concordant alignment. The proportion of discordant read alignments in a dataset is not trivial and sometimes it is hard to distinguish concordant alignments from discordant alignments (largely depending on the parameter setting). Therefore it is better to let the aligner do its best to find the best possible alignments no matter they are concordant or discordant, as long as the mapping quality is good (eg. number of mismatches in each mate is acceptable).

The bowtie2 parameter '--score-min' lets user set a function to specify the minimum alignment score. Only read length is considered in this function. Rsubread aligners do not support this filtering. But the filtering supported by Rsubread aligners, such as minimum number of votes and maximum number of mismatches, should be able to achieve similar effect. Also the alignment score reported by Rsubread aligners considers more information such as base Phred scores and number of mismatches, in addition to read length. Downstream tools often need a full spectrum of alignment scores and therefore it is better not to filter mapped read by alignment scores at this mapping stage.