A model of pulldown alignments from SssI-treated DNA improves DNA methylation prediction

Area of Science:

• Epigenetics
• Genomics
• Bioinformatics

Background:

• Methyl-CpG binding domain (MBD) protein pulldown followed by high-throughput sequencing is a standard method for DNA methylation analysis.
• Accurate quantification of DNA methylation from MBD-seq data typically requires control data from SssI-treated samples.

Purpose of the Study:

• To develop a computational model to predict MBD pulldown reads from SssI-treated DNA.
• To assess the effectiveness of this model in improving DNA methylation level estimations.

Main Methods:

• A model of expected MBD pulldown reads was constructed based on prior MBD2 binding characterizations.
• The BayMeth program was utilized to evaluate the model by substituting calculated SssI control data for observed data.
• Model performance was compared against standard BayMeth analysis and validated on external datasets.

Main Results:

• BayMeth using the modeled SssI control data outperformed BayMeth using observed SssI control data for methylation prediction.
• The modeled control data improved predictions on both 100 bp and 10 bp windows.
• The model demonstrated comparable performance to observed control data on an external dataset, requiring only adjustment of average fragment length.

Conclusions:

• The developed MBD pulldown alignment model enhances DNA methylation predictions.
• The model improves accuracy without necessitating additional SssI control experiments.
• This approach offers a more efficient and accurate method for MBD-seq data analysis.