Bowtie2 with secondary alignment option produces multiple mapping with offset and varying alignment scores

Question

[polag01]

Please does anyone have a way to resolve the multiple alignment from bowtie2 to produce alignments with only the best alignment score?

A hit at a locus could be shifted about 1bp with another alignment reported but with a different alignment score. The same hit could be shifted by 6bp with another different alignment score.

This I believe is happening due to the -secondary switch to output secondary alignments. I need the secondary alignments to know where the copies of the reads are. I want only the high quality alignments for a read at a locus without having the ones at locus+N1, locus+N2 etc.

An illustration is shown in the figure below.

The alignment at locus 2 ends at 688, shifting the read by 1 base, there is another alignment with a different alignment score.

I should state that I am aligning to a chromosome with highly repetitive DNA. I am using bowtie2 due to its higher sensitivity.

Any help would be highly appreciated.

Thanks in advance.

[rfran010]

I am a little unclear on your question.

What is the '-secondary' switch?

Overall, you want the secondary alignments, but you want bowtie2 to know not to give you a secondary alignment if it overlaps with another alignment from the same read?

Can I ask why you need the secondary alignments? If you want to know where in the chromosome the copies of the read are, wouldn't it just be better to analyze the chromosome directly to annotate all the copies?

[polag01]

Thanks for your reply. The --secondary=yes switch is what I referred to. Yes, I want secondary alignments because they are repeat DNA sequence but I thought I wouldn't get all the copies without asking for a report of the secondary alignments.

[rfran010]

And you're sure the +1 alignments are from the same read?

Yes, I'm not familiar with the secondary switch you are referring to regarding to bowtie2. If I google it, I only see this option for minimap2.

I am confused because for bowtie2 to output multiple alignments for the same read I only know of -k and the behavior of this option is to search for the given number of distinct alignments.

[polag01]

Yes, the +1 alignments are from the same read, this is clear from the image I posted earlier.

You are quite right, it was a grave mix-up on my end. The switch indeed is from minimap2 (I had been trying to compare the two aligners), so sorry for the confusion!

[polag01]

The bowtie2 parameters I used were:

bowtie2 --local -a -N 1 -p $threads \ -f -x $refIndex -U $inFile2 \ --very-sensitive-local -S $outputDir"/"$sample"_"$inputTAG".sam"