What does fully overlapping supplementary alignments mean?

Question

[njornet]

Hello everyone,

I have some human nanopore data that I aligned using minimap2 version 2.28-r1209. In the BAM file there are reads with supplementary alignments that almost fully overlap in the forward and reverse strands. I know supplementary alignments can be chimeric reads but it can't be my case as I said they are fully overlapping. Is this some kind of error or can supplementary alignments mean something else?

Answer 1

[Michael]

RepeatsRepeatsRepeats Low complexity

Repetitive sequences cause most funny alignment results. I would always check that first. IF the sequence is palindromic in addition, it would satisfy the condition of aligning to the forward and reverse strand. The easiest sequence to satisfy that would be long stretches of AT or GC. You can always look at your aligned sequences more closely in a genome browser like IGV if you want to be certain. In most cases, you can simply ignore those alignments. All depends on the question you want to answer by re-sequencing. If you want to do structural variant analysis, the pipeline should deal with those alignments.

[njornet]

Sorry, I forgot to mention I already checked and I'm pretty sure it's not a repetitive region nor palindromic, I have tons of reads that do this but one example aligns to chr7:117,588,913-117,590,568 (-) and chr7:117,588,923-117,590,566 (+).

[Michael]

If you are really interested in those sequences, you should extract some of the reads and the aligned region, align them in Jalview or another tool, and inspect the alignment visually. If we assume there is no bug in minimap and the alignment is real, though, you are possibly just going to wind up in something that carries no real relevance. Also, if 1) the align is correct and 2) the same sequence aligns completely to forward and reverse strand, it is a palindrome by definition, just maybe a very long one. You should also check how much query coverage there really is within the alignment.