search

Shians / NanoMethViz

https://shians.github.io/NanoMethViz/
Apache License 2.0
23 stars 1 forks source link

Can I directly use log_m_ratio matrix to do methylation differential analysis? #28

Closed YingziZhang-github closed 1 year ago

YingziZhang-github commented 1 year ago

Hi Shian,

Following the guideline in R help, I used the conversion function bsseq_to_log_methy_ratio to convert from my BSseq object bsseq to a matrix of log-methylation ratios log_m_ratio. I understand that the log_m_ratio matrix can be used with the plotMDS or plotPCA function.

If I want to do differential methylation analysis and to distinguish differential methylated loci and loci without significant methylation level changes within one sample and in different treatment conditions. Can I directly use the values in log_m_ratio matrix? Do you have suggestions what statistical method should I use?

Thank you very much for your time. Looking forward to your reply!

Yingzi

Shians commented 1 year ago

The methylation ratios shouldn't be used for differential methylation analysis, because the ratios don't capture the coverage which is important for statistical methods. For example we are more confident about a loci that has 0.5 methylation if we captured 50 methylated CG and 50 unmethylated CG rather than 1 and 1. I recommend using the BSSeq object directly with DSS or bsseq. I believe DSS can work when there are no replicates within conditions, not sure about bsseq.

YingziZhang-github commented 1 year ago

Hi Shian,

Thank you for the explanation. I am moving on with DSS instructions for further differential analysis. Many thanks.

Yingzi