Guidance on sample read depth for m6A comparisons across samples

[hannalee809]

Thank you for developing m6Anet, which has been incredibly helpful for my work. I'm currently using it to compare m6A methylation profiles across experimental groups, but I’ve noticed that the number of mapped reads varies significantly between samples (ranging from ~20,000 to 560,000). As expected, the number of high-confidence sites (probability > 0.9) appears to correlate with read depth.

I was wondering if you could share any recommendations for ensuring accurate comparisons across samples, particularly regarding normalization strategies or best practices for handling differences in read coverage. Please let me know if you need more information. Thank you!

[yuukiiwa]

Hi Hanna (tagging you here @hannalee809),

We have a tool called xpore for differential modification detection, which performs a z-test that allows fair comparisons across samples with varying read coverages.

Thanks!

Best wishes,
Yuk Kei