Circulating cell-free DNA (cfDNA) methylation sequencing is becoming a key technology for liquid biopsy and early cancer detection. By analyzing methylation patterns in plasma DNA fragments, researchers can identify tissue-of-origin (TOO) signals and detect cancer-associated epigenetic changes.
Figure 1. Concept of cfDNA methylation sequencing and nucleosome-associated fragmentation.
However, one important technical factor is often overlooked: the library preparation strategy.
Two main approaches are commonly used:
- Double-stranded DNA (dsDNA) library preparation
- Single-stranded DNA (ssDNA) library preparation
These methods can lead to different methylation measurements and biological interpretations.
1. dsDNA vs ssDNA Library Prep: The Core Difference
The biggest difference between dsDNA and ssDNA library prep lies in how fragmented DNA ends are processed.
dsDNA library prep: Requires enzymatic end repair, which adds unmethylated nucleotides and reduces CpG methylation at fragment ends.
ssDNA library prep: Skips this step, preserving native ends and true methylation signals.
2. Experimental Workflow
Sample Preparation
Plasma cfDNA was extracted from whole blood. DNA concentration was measured using a Qubit fluorometer, and fragment size distribution was assessed using the Agilent Technologies TapeStation system.
Libraries were constructed from 1 ng cfDNA using either double-stranded (dsDNA) or single-stranded (ssDNA) methylation library preparation methods.
Library Preparation Overview
|
Parameter |
dsDNA Library |
ssDNA Library |
|
Library strategy |
Double-stranded DNA |
Single-stranded DNA |
|
Recommended kit |
Yeasen-dsDNA Methyl Library Prep Kit for Illumina™ (Cat#12214ES) |
Methyl-seq ssDNA Library Prep Kit |
|
Input DNA |
1 ng cfDNA |
1 ng cfDNA |
|
PCR cycles |
11 cycles |
11 cycles |
Sequencing and Data Processing
Libraries were sequenced on the Illumina NextSeq 2000 platform using 2 × 66 bp paired-end reads.
Bioinformatics Pipeline
|
Step |
Tool |
Purpose |
|
Adapter trimming |
SeqPrep2 |
Remove adapter sequences |
|
Alignment |
bwa-meth |
Map reads to reference genome |
|
Duplicate removal |
Picard |
Mark PCR duplicates |
|
Methylation calling |
MethylDackel |
Extract CpG methylation |
Filtering criteria:
- Minimum mapping quality ≥ 20
- Minimum sequencing depth requirement
Nucleosome-Associated Methylation Analysis
Nucleosome positioning information was obtained from NucPosDB and converted to 1 bp dyad resolution.
Using bedtools, the distance between CpG sites and nucleosome centers was calculated to evaluate methylation patterns relative to nucleosome positioning.
Tissue-of-Origin (TOO) Inference
Tissue-of-origin prediction used a methylation reference atlas from Loyfer et al., obtained from the Gene Expression Omnibus dataset GSE186458.
The reference included 13 cfDNA-relevant tissues, such as endothelial cells, granulocytes, and hepatocytes.
TOO Analysis Strategy
|
Parameter |
Description |
|
Algorithm |
Non-negative least squares (NNLS) |
|
Training/Test split |
70% / 30% |
|
Iterations |
100 |
|
Evaluation metrics |
R² and Mean Squared Error (MSE) |
End-Trimming Correction Analysis
To evaluate correction of methylation bias, dsDNA sequencing reads were trimmed by 20 bp and 40 bp from read ends.
Methylation levels and tissue-of-origin prediction accuracy were then reanalyzed.
3. Results
Two cfDNA samples were used to generate paired ssDNA and dsDNA libraries. Key findings are summarized below.
Table 1. Basic Characteristics of the Two Library Preparation Methods
Figure 1. CpG Methylation Levels Across Read Length
3.1 dsDNA Libraries Show Methylation Bias
|
Observation |
dsDNA Library |
ssDNA Library |
|
Read-level methylation pattern |
CpG methylation decreases after ~40 bp |
Uniform across reads |
|
Reverse read 3′ end |
Sharp methylation loss |
Stable methylation |
|
Nucleosome-associated pattern |
Strand-specific hypomethylation |
Symmetric methylation |
The bias in dsDNA libraries is consistent with end-repair filling of 5′ overhangs with unmethylated cytosines.
3.2 dsDNA Libraries Show Lower Global Methylation
|
Metric |
dsDNA |
ssDNA |
|
Global CpG methylation |
Reduced |
Higher |
|
Estimated difference |
~5–7% lower |
Baseline |
This reduction overlaps with nucleosome-associated methylation valleys, amplifying measurement bias.
3.3 ssDNA Libraries Improve Tissue-of-Origin Prediction
|
Metric |
dsDNA |
ssDNA |
|
R² |
Lower |
Higher |
|
MSE |
Higher |
Lower |
|
Informative CpG subsets |
Reduced signal |
Strong signal |
The improvement is particularly evident for high-information CpG sites:
methylation < 0.2 or > 0.8
methylation < 0.1 or > 0.9
For uninformative unmethylated CpGs, both methods showed similar performance.
3.4 End Trimming Partially Corrects dsDNA Bias
|
Trimming length |
Effect |
|
20 bp |
Partial correction |
|
40 bp |
Methylation closer to ssDNA |
After trimming:
- R² increased
- MSE decreased
However, trimming also introduces limitations:
- Reduced unique sequencing coverage
- Loss of fragment end coordinates
- Reduced ability to perform fragmentomics analysis
4. Conclusion
|
Key Finding |
Impact |
|
End-repair in dsDNA library prep introduces unmethylated cytosines |
Causes methylation underestimation |
|
ssDNA library prep avoids end repair |
Preserves native fragment ends |
|
ssDNA libraries retain authentic methylation signals |
Improves cfDNA methylation analysis |
ssDNA library preparation better preserves native cfDNA methylation signals. Meanwhile, dsDNA workflows remain widely used due to their simplicity, robustness, and scalability for routine sequencing applications.
Related Product
|
Product Name |
Cat. No. |
Application |
|
HieffTM Superfast DNA Methylation Bisulfite Kit |
12228ES |
Column-based DNA methylation conversion |
|
HieffTM Mag Superfast DNA Methylation Bisulfite Kit |
12229ES |
Magnetic bead-based DNA methylation conversion |
|
Hieff NGSTM Methyl-seq ssDNA Library Prep Kit for Illumina V2 |
12221ES |
SsDNA methylation library prep for Illumina |
|
Hieff NGSTM Methyl-seq ssDNA Library Prep Kit for MGI V2 |
12226ES |
SsDNA methylation library prep for MGI |
|
12214ES |
DsDNA methylation library prep for Illumina |
|
|
12218ES |
Enzyme-based methylation fragmentation |
|
|
Hieff CanaceTM Uracil+ High-Fidelity DNA Polymerase Mix |
12928ES |
Post-bisulfite PCR amplification |
Reference
[1] GROTH T E, MISHIN A A, RAO V, et al. End-repair causes methylation underestimation in cell-free DNA sequencing libraries[EB/OL]. bioRxiv Preprint, 2025[2025-12-17]. https://doi.org/10.64898/2025.12.15.694439.