minimap2: aligning multiple regions to a read in a fastq

Question

[vinaykusuma]

This is my command

$ minimap2 -t 4 -ax sr -p0 -N 6 -k 10 transciptome.fa reads.fastq > aligned_reads.sam

trancriptome.fa looks like this, it has more than 500 headers, I have added some here for example.

>12324_A  
TGCGGCCAATGTTTGTAATCAGCCCATAAATCCCAAGGAAATTTTGGGGAC  
>43354_B     
TGCGGCCAATGTTTGTAATCAGCGACCTAGGACCAAGGAAATTTTGGGGAC   
>42343_C   
TGCGGCCAATGTTTGTAATCAGGCGATTGTAACCAAGGAAATTTTGGGGAC    
>23435_D    
TGCGGCCAATGTTTGTAATCAGCACTTAGGAGCCAAGGAAATTTTGGGGAC

one of these transcript is mapped to one read in fastq but I expect multiple mapping.
Suppose, lets take Read 1 from fastq. Suppose it can match 1234_A and 43354_B but
I see in my .sam that it matches to only either one of them. I want to know all possible transcripts
matching a read.

Thanks.

[GenoMax]

If your reads are longer than the transcriptome (which looks suspiciously small in length) then you could flip the query/database. You should check on the secondary alignments. Looks like you are asking for 6 possible alignments but don't have --secondary=yes.

[vinaykusuma]

Hello, These are not transcriptome actually. I am trying to get count of each sequence in transcriptome.fa that is present in my reads.fastq So, my plan was to align these sequences to read.fastq and then count the alignments. I am not sure if that will work.

[GenoMax]

Why not align your reads to transcriptome.fa and then use samtools idxstats get the counts. You can even use salmon to get an estimate of expression instead of doing alignments. Is there a specific reason you are using minimap2? If you have regular Illumina reads use bwa mem or one of the normal NGS alingers.

[vinaykusuma]

The reason behind using minimap is that it gives me flexibility to switch to long reads option from short reads if required. Can I use salmon when I have long reads? thanks for replying.

[Rob]

Yes, you can use salmon when you have long reads — however, you should use alignments generated with a long-read oriented aligner rather than salmon's inbuilt selective-alignment (which has not been thoroughly tested with long reads yet). There is a pipeline for long read quantification that makes use of salmon available here. This is useful, because there are certain flags that one should use when doing long-read quantification with salmon.

[vinaykusuma]

Thanks Rob, I will surely check out and try it. Thanks again.