QV score lower post dorado polish

[HananAlmulla]

Issue Report
I assessed the Q score of my assembly using YAK. the QV score of my draft assembly is 55 while that of the polished assembly using dorado polish is much lower (35).

Steps to reproduce the issue:

1. convert POD5 files to bam file using dorado basecaller (v0.8.2) and SUP v5.0 model
2. perform assembly using Hifiasm including recent --ont option (v0.23.0-r691)
3. run dorado aligner
4. run 'dorado polish` (v0.9.0)
5. run yak count (v0.1-r56)
6. run yak qv on both the draft and the polished asssembly

Run environment:

• Dorado version: v0.8.2 for basecalling and v0.9.0 for polishing
• Operating system: Ubuntu 22.04.5 LTS
• Hardware (CPUs, Memory, GPUs): AMD® Ryzen threadripper pro 7985wx 64-cores × 128 - 256.0 GiB mem - NVIDIA GPUs
• Source data type: POD5 files
• Source data location: on device
• Details about data (flow cell, kit, read lengths, number of reads, total dataset size in MB/GB/TB):
  • flowcell: PromethION R10 flowcell
  • kit: Native Barcoding Kit 24
  • read length: n50 is 4251 bp
  • number of reads: 14222718

[svc-jstone]

Hi @HananAlmulla,
Thanks for reporting this! We would like to look into it more to understand the issue.
Is it possible that you share a small part of your data (e.g. a single smaller draft contig which exhibits this issue and accompanying aligned reads) so we can have a deeper look?