A computational DNA methylation method to remove contaminated DNA from spent embryo culture medium for noninvasive preimplantation genetic testing

Background In the last fifty years, assisted reproductive technology (ART) has achieved remarkable breakthroughs, culminating in the birth of 12 million infants. At the heart of ART success is preimplantation genetic testing (PGT), which enables the detection of chromosomal anomalies, single-gene disorders, and structural rearrangements, enhancing embryo selection and mitigating genetic risk. However, current PGT methods, including trophectoderm (TE) biopsy, face limitations such as challenges related to convenience and potential impacts on embryonic health. In this evolving field, noninvasive PGT (niPGT) has emerged as a promising alternative. By analysing cell-free DNA (cfDNA) in spent embryo culture medium (SECM), niPGT offers a less intrusive approach. However, maternal DNA contamination within SECM remains a marked barrier to its clinical application as underscored by our research and other studies. There is an urgent need for innovation and optimisation in niPGT methodologies. Methods We developed a computational algorithm to eliminate contaminated nonembryonic DNA from spent embryo culture medium. The rationale is based on the phenomenon that the DNA methylation level of a mammalian preimplantation embryo reaches its minimum at the blastocyst stage during a global DNA demethylation wave. Therefore, selecting hypomethylated reads is expected to enrich blastocyst DNA over nonembryonic DNA. To investigate this, we retrieved single-cell-resolution DNA methylation data from oocytes (n = 33), inner cell masses (ICMs, n = 74), TEs (n = 71) and sperm cells (n = 21), bulk DNA methylation data from cumulus cells, and DNA methylation data from SECM samples (n = 194) from our previously published database, and conducted a comparative analysis of DNA methylation patterns among them. Then, we constructed a decontamination algorithm based on single read and applied it to remove contamination originating from cumulus cells, polar bodies, and sperm cells. Findings By selecting hypomethylated reads, we successfully enriched blastocyst DNA over DNA originating from cumulus cells, polar bodies and sperm (enrichment factors = 4, 1.2, and 2.5, respectively). By testing simulated SECM samples, the method demonstrated a substantial reduction in the false-negative rate even with up to 75% cumulus cell contamination. In real clinical SECM samples, the method improved aneuploidy detection sensitivity at a cumulus cell contamination ratio of 50%. Interpretation Our study introduces a novel computational strategy for reducing nonembryonic DNA contamination, thereby enhancing aneuploidy detection sensitivity in SECM cfDNA methylation analyses. In combination with DNA methylation methodologies, this approach holds considerable promise for advancing niPGT applications in ART.