Fig 1.
Summary flowchart of the methods used in this study.
*If a pre-transplant recipient sample is not available, leukocyte-derived genomic DNA or DNA from other sources can be fragmentase-treated and used as a substitute pre-transplant sample. **Although SNPs are categorized into three groups based on ALT ratios, only two groups—ALT 0.9–1.0 and 0.0–0.1—are used in the subsequent analysis.
Fig 2.
Changes in alternative allele ratio in a mixing experiment using fragmentase-treated genomic DNA mimicking recipient cfDNA.
Fragmented genomic DNAs from two individuals were either left unmixed (A) or mixed at 1% (B) or 3% (C). These samples underwent capture hybridization and NGS sequencing to determine alternative allele ratios. In each plot, SNPs are sorted in descending order of alternative allele ratio. In panels (B) and (C), mixed samples are shown in orange and unmixed controls in dark blue. Reference alleles are denoted by A and a; alternative alleles by B and b. Uppercase letters indicate recipient-derived alleles, while lowercase letters indicate donor-derived alleles. The corresponding genotype combinations are illustrated in the figure.
Fig 3.
Estimation of dd-cfDNA fraction using k-means++ clustering.
(A) Scatter plot of alternative allele ratio differences for all SNPs in which the recipient genotype is BB (n = 300). The data were grouped into three clusters using k-means++ clustering, corresponding to BB/bb, BB/ab, and BB/aa. (B) Scatter plot of alternative allele ratio differences for all SNPs in which the recipient genotype is AA (n = 300). The data were grouped into three clusters using k-means++ clustering, corresponding to AA/aa, AA/ab, and AA/bb.
Fig 4.
Validation of dd-cfDNA estimation using genomic DNA mixing.
(A) Correlation between the actual mixing ratios of fragmentase-treated genomic DNA and the estimated dd-cfDNA fractions across the full range (0–100%). A strong linear relationship was observed (r² = 0.9987, y = 0.9953x + 0.0059). (B) Focused view of dd-cfDNA fractions within the clinically relevant range (0–10%), demonstrating continued high correlation (r² = 0.9956, y = 1.0006x + 0.0032). (C) Estimation of dd-cfDNA fractions without using pre-transplant reference data. SNPs were grouped based on observed alternative allele ratios, and clustering was applied within each group. The estimated values were highly correlated with the actual ratios (r² = 0.9973, y = 0.8574x + 0.0032), though with a slight underestimation (“S4 Sheet 100%_wo_pre-data” in the Supporting Information).
Table 1.
Assessment of Clustering-Based dd-cfDNA Estimation by In Silico Downsampling.
Table 2.
Comparison of dd-cfDNA ratios estimated using clustering-based methods and direct genotypic calculations.
Fig 5.
Comparison of clustering-based estimation with direct genotype-based calculation.
Estimated dd-cfDNA ratios obtained using the clustering-based method with pre-transplant SNP data (orange:clustering_pre) and without pre-transplant SNP data (skyblue:clustering_wo) are compared to directly calculated values based on known donor and recipient genotypes (green:direct method). Each panel shows the time course of dd-cfDNA levels following kidney transplantation: (A) unrelated donor–recipient pairs, (B) sibling pairs, (C) parent–child pairs.