Slam-Seq Conversion Rate per Read

Question

[Daniel]

Hello,

I've analyzed slam-seq data using the relatively older nf-core slamdunk pipeline. I've been told to proceed with the data because it's from a very high-impact publication, but I am confused on how the data passes QC. I thought that T>C conversion plots are very important to tell whether an slam-seq was successful, and to me, it seems that the time points here do not at all correlate with the conversion rates. Is there something I am missing? Thank you!

[michael.ante]

Try to switch to percentage view of this graph. Stacked bar charts with different heights can sometimes be misleading.

Answer 1

[Kevin Blighe]

The T>C conversion plots from nf-core/slamdunk can look flat across time points if they're showing raw counts instead of normalized percentages—check by dividing T>C events by total T bases or reads per sample, which should reveal the expected rise from low background (~0.1-0.5% at 0h) to higher rates (1-5%+) in later time points for successful 4sU labeling. If still off, subtract the 0h baseline, relax multi-TC filters for QC, or verify UTR/CDS breakdowns, as the high-impact pub likely normalized internally; updating to the latest pipeline version could help with better stats too.

Goodnight