Bug report

[tcztzy]

It seems to be a typo at following line.

HelixerPost/helixer_post_bin/src/analysis/hmm.rs

Line 880 in 6f8ab1c

HmmPrimaryState::Coding1 => (pred.get_coding_phase2_penalty(), pred.get_coding_phase2_neg_log_prob()),

[felicitas215]

Hi, thank you for noticing this. This was indeed a bug that was fixed in a side-branch. Since it had a very minor influence on the predictions in the tests we conducted, we opted to wait with the fix and some other improvements to HelixerPost until the Helixer paper was published. We expect to release updated version of HelixerPost soon.

[colindaven]

Hi @felicitas215 I saw this too when attempting to parallelize helixer_post with Claude AI. I fixed, then unfixed this to match (this) reference implementation. And now I see it is a known unfixed bug.

The interesting thing is that it makes a major difference in my test data of two maize organelles.

• This github reference implementation - 171 genes
• My parallel implementation (this bug was fixed) - 237 genes
• Another container built on the latest release (Oct 22 2025 data) - 239 genes.

So I'm not sure what I should really build from and compare against.

Would you recommend I compare against the Oct 22 2025 release ? Or against the latest commit ?

Thanks and best wishes,
Colin