Assemblers for Very Short Products using Long Reads

[JW_REGN]

I have some ONT data for a mixed pool that includes one very short product ~500 bp. My standard de novo assembly pipeline using Flye fails to assemble these reads due a parameter that sets a hard minimum on overlap sizes that excludes this small product. I've been running through a list of long-read assemblers on the subset of reads in the 400-600 bp range but all the assemblers either have a similar parameter issue or choke on the contig construction. Closest one I've been able to get to completion is canu which still fails at unitigging. A general problem seems to be that many of the reads cover the full sequence in a single read and thus most of my reads are contained fully by other reads and get filtered out leaving nearly no data for contig construction.

Anyone have any experience getting successful assemblies from long-read data for such short products?

[incidental]

You might want to try relaxing the overlap/min-read thresholds even further or testing assemblers that are a bit more tolerant of low coverage/short targets. For ~500 bp products, a lot of de novo pipelines are kind of optimized against fragments that small and end up filtering them out as noise or contained reads driving directions.

Also, have you tried pulling only the reads that map to a close reference (if one exists) and assembling that subset separately? Even a lightweight reference-guided approach or targeted assembly might help if the full pool is drowning out the short product