Does ccs dislike low-complexity regions?

Low-complexity comes in many shapes and forms. A particular challenge for ccs are highly enriched tandem repeats, like hundreds of copies of AGGGGT. Prior ccs v5.0, inserts with many copies of a small repeat likely not generate a consensus sequence. Since ccs v5.0, every ZMW is tested if it contains a tandem repeat of length --min-tandem-repeat-length 1000. For this, we use symmetric DUST and in particular the sdust implementation, but slightly modified. If a ZMW is flagged as a tandem repeat, internally --disable-heuristics is activated for only this ZMW, and various filters that are known to exclude low-complexity sequences are disabled. This recovers most of the low-complexity consensus sequences, without impacting run time performance.