Dear herro authors,

For my particular use case I find that the minimap2 all-v-all overlap stage results in a large fraction of input reads being discarded before the inference stage (2828 reads in overlap input, 620 represented in overlap alignment output).

Please can I ask how the ava overlap parameters have been chosen in the create_batched_alignments.sh script?:

minimap2 -K8g -cx ava-ont -k25 -w17 -e200 -r150 -m2500 -z200 -f0.005 --dual=yes

I am using herro for error correction across reads from a small fraction of the human genome (~400kbp locus) which is known to contain a large segmental duplication. The low number of herro output reads resulting from loss at the overlap stage, means that there are some regions in the locus with extremely low coverage. Although I can adapt the minimap2 parameters to improve this (in particular, allowing k, w, and f to take default -X ava-ont values), I would like to understand whether this would be a bad idea for herro inference stage.

Any insights gratefully received, thank you.