How can I use the output of histone modifications prediction to interpret the different/significant histone combination in each peak?

Help & Support
1

Hi AlphaGenome team,

Thank you for developing this amazing tool.

I’m currently using it to predict histone modifications for a list of peak regions of interest. As we know, histone modifications always act in combination to regulate chromatin accessibility. I would like to identify which histone-mark combinations are significantly enriched or characteristic for each peak set.

Given that the predicted histone-signal matrix resembles a pseudo-read-count matrix, where each value can be interpreted as a depth-like signal for a specific peak and histone mark. I’m wondering whether you have suggestions or recommended approaches for identifying meaningful or significant histone-modification combinations from this matrix.

Any guidance on analytical strategies or best practices would be greatly appreciated.

Thank you very much!

Xiaohuan

2

Hi There!

Thanks for reaching out.

To identify meaningful histone-modification combinations from your pseudo-count matrix, have you considered using unsupervised chromatin state learning tools like ChromHMM (requires data binarization) or Spectacle (using continuous signals) to segment peaks into distinct functional states?

Kind regards,
Tumi