DNA 5-methylcytosine detection and methylation phasing using PacBio circular consensus sequencing

Area of Science:

• Genomics
• Epigenetics
• Bioinformatics

Background:

• Long-read sequencing technologies like PacBio CCS and nanopore sequencing show promise for detecting DNA 5-methylcytosine (5mCpG).
• Existing PacBio CCS methods for 5mCpG detection lack accuracy and robustness, particularly in complex genomic regions.

Purpose of the Study:

• To develop a novel deep-learning method, ccsmeth, for accurate and robust detection of DNA 5mCpGs using PacBio CCS reads.
• To create a Nextflow pipeline, ccsmethphase, for haplotype-aware methylation detection from CCS data.

Main Methods:

• Trained ccsmeth using PacBio CCS reads from polymerase-chain-reaction treated and M.SssI-methyltransferase treated human DNA.
• Evaluated ccsmeth's performance on 5mCpG detection at single-molecule resolution using long CCS reads (≥10 Kb).
• Developed ccsmethphase pipeline for haplotype-aware methylation analysis and validated it using a Chinese family trio.

Main Results:

• ccsmeth achieved 0.90 accuracy and 0.97 Area Under the Curve for single-molecule 5mCpG detection.
• Genome-wide site-level analysis showed >0.90 correlation with bisulfite and nanopore sequencing using only 10× CCS reads.
• ccsmethphase successfully detected haplotype-aware methylation patterns.

Conclusions:

• ccsmeth provides a robust and accurate deep-learning approach for DNA 5mCpG detection using PacBio CCS long reads.
• ccsmethphase enables sensitive haplotype-aware methylation analysis, advancing epigenetic studies.
• These tools offer significant improvements for DNA methylation analysis, especially in challenging genomic areas.