# A SNP-based capture and clustering workflow to assess donor-derived cell-free DNA in transplantation

## TL;DR

This paper introduces a new method using SNPs and clustering to accurately measure donor-derived cell-free DNA in transplants, which could help detect organ rejection early.

## Contribution

A novel SNP-based capture and clustering workflow for estimating donor-derived cell-free DNA with high accuracy and robustness.

## Key findings

- The method achieved an r² of 0.9987 with 0% mixture samples across the full 0–100% range.
- Clustering-based estimates showed high concordance with direct calculations in kidney transplant recipients.
- The method remains accurate even with reduced input and when pre-transplant data are unavailable.

## Abstract

Measurement of donor-derived cell-free DNA (dd-cfDNA) enables early, non-invasive monitoring of transplanted organs, including rejection detection. We developed a method to estimate dd-cfDNA ratios using capture hybridization of 300 SNPs, next-generation sequencing (NGS), and clustering analysis. Validation was conducted using simulated mixtures of fragmented genomic DNA from two individuals (0–100%). dd-cfDNA ratios were estimated via clustering, with and without 0% mixture samples to simulate the presence or absence of pre-transplant recipient plasma. When 0% samples were included, estimation achieved an r² of 0.9987 across the full 0–100% range; without them, r² remained high (0.9973) in the clinically relevant 0–10% range. The robustness of the method was further demonstrated by in silico downsampling. MAEs with 0% samples were 0.823%, 0.766%, and 0.702% at full, 50%, and 25% read depths, respectively (0–100% range). For the 0–10% range, MAEs were 0.333%, 0.300%, and 0.467% with 0% samples, and 0.413%, 0.367%, and 0.503% without them. These results indicate that the method maintains high accuracy even under reduced input and when pre-transplant data are unavailable. We also compared clustering-based estimates with direct calculations from kidney transplant recipients, where donor and recipient SNP genotypes were known. The concordance correlation coefficient (CCC) from day 0 to day 28 post-transplantation was 0.9887 and 0.9316 for unrelated pairs with and without pre-transplant data, respectively. For sibling pairs, CCCs were 0.9923 and 0.9675; for parent–child pairs, the CCC was 0.9831 with pre-transplant data. CCC was not calculated for parent–child pairs without pre-transplant data due to limited samples (<10%, n = 3). These findings demonstrate high concordance, accuracy, and robustness of our clustering-based dd-cfDNA estimation method and support its potential utility in clinical transplantation settings.