Fetal fraction amplification within prenatal cfDNA screening enables detection of genome-wide copy-number variants at enhanced resolution.

Abstract

Clinically significant copy-number variants (CNVs) occur in 1% to 2% of pregnancies and are difficult to detect via prenatal cell-free DNA (cfDNA) screening because of the low fraction of fetal-derived cfDNA in maternal plasma. Here, we use fetal fraction amplification (FFA) and improved computational algorithms to enhance the resolution and sensitivity of CNV detection. We implemented and characterized the performance of a hidden Markov model that identifies fetal CNVs. This CNV caller was analytically validated on 117 FFA samples, including 57 fetal-CNV-containing samples, and applied retrospectively to a cohort of more than 300k patient samples. Our assay was concordant with orthogonal testing and detected fetal CNVs ≥5 Mb with estimated aggregate sensitivity and specificity of >95.1% and >99.7%, respectively. The resolution of CNV detection was fetal fraction dependent, but 97.2% of samples reached ≥5-Mb resolution. Overall, CNVs ≥5 Mb were found in 1 in 500 pregnancies. FFA improves the sensitivity and resolution of CNV detection in prenatal cfDNA screening, allowing accurate detection of fetal CNVs as small as 1 Mb. Using our approach, we found that clinically significant fetal CNVs were detected more frequently than the common trisomies 13 and 18 that are recommended as part of guideline-based screening.