Handling differences between short and long reads

[LehmannN]

Hi,

I wondered if you ever tested portcullis on both short reads and long reads, and if yes, how does it affect the results. For the experiment I am working on, we have short reads that are very short (56pb long) so it might be relevant to use only the long reads to infer the splice junctions. Data from long and short reads are from the same cell types. Do you have any advice for this type of approach ?

[lucventurini]

Hi @LehmannN

Unfortunately Portcullis is not ready to be used with long reads. I have worked a bit on it last year, but other projects took priority.
Anedoctally, I would recommend trying desalt for aligning your long reads. In my quick tests last year it seemed to perform really well. Another good alternative, if you have IsoSeq, is - again in our experience - GMAP.

[LehmannN]

Hi @lucventurini,
I aligned my Nanopore long-reads with Minimap2 and then used the Minimap2 output BAM file as input of Portcullis but I still didn't have time to analyse the output (it looks quite complete though). I will also try with desalt then !

[lucventurini]

Yes, I highly recommend using DeSALT instead of minimap2. Basically DeSALT does what HISAT2 does with short reads - align once, find the best junctions, realign using those junctions as a guide.
Internally it actually uses minimap2 for aligning reads.
In my very quick and non scientific test last year, the accuracy went marvelously up - like, 30-40% better using DeSALT. So for your project I would highly recommend going that route, especially because mikado cannot salvage a bad intronic alignment!

[LehmannN]

Ok, thanks a lot for your feedback ! I will let know how it goes with DeSALT.