Tailoring resolution for PGT-A and PGT-SR by modifying NGS sample throughput

Abstract

Unlike traditional DNA analysis platforms used for Preimplantation Genetic Testing (PGT), Next Generation Sequencing (NGS) platforms provide a highly flexible and customizable workflow for the detection of segmental copy number variations (CNVs) offering dynamic resolution. However, greater resolution is achieved at the expense of sample throughput. Herein, we aim to define the level of resolution that can be achieved with RHS’ PG-Seq with different NGS throughput. PG-Seq NGS files representing different sample throughput scenarios and containing 100Kb to 5Mb CNVs were created in silico then analysed with PG-Seq software. NGS files reflecting the expected total read count per sample from a single Illumina MiSeq run with throughput from 1-96 samples were created in silico from PG-Seq NGS files. Random sampling was used to create files with total read counts of 250,000 (96 samples), 500,000 (48 samples recommended for PG-Seq), 1 million (24 samples), 3 million (8 samples), 12 million (2 samples) and 24 million (1 sample). Artificial CNVs, comprising gains and losses ranging from 100Kb to 5Mb were added across chromosomes 1, 5, 15, 17, 19, and 22. PG-Seq software was used to call CNVs. Detection of CNVs of each size are detailed in the table below for each corresponding total NGS read count. High resolution CNV calling, for the detection of germline or de novo 100Kb segmental CNVs, required > 12 million reads, or no more than 2 samples per MiSeq run. A standard 48 sample PG-Seq run reliably detected segmental CNVs of 5Mb and should theoretically be able to detect de novo CNVs of this size. Using a higher throughput of 96 samples per run, reliable detection of CNVs should be possible down to 10Mb; while germline segmental CNVs should be detected at this low resolution, there is a risk of false positives if calling de novo segmental CNVs. In silico analysis of sequencing data highlights the technical limitations and financial implications for the detection of small segmental CNVs by NGS. If greater resolution is required, throughput needs to be reduced. Decisions regarding increased resolution and the need to validate workflows using cell line or clinical samples would best be made on a case-by-case basis, should the need arise.Tabled 1Percentage of CNVs of different size detected for each read count scenario.CNV detection % (number detected/expected)250,000 reads (96 samples)500,000 reads (48 samples)1 million reads (24 samples)3 million reads (8 samples)12 million reads (2 samples)24 million reads (1 sample)Size of CNV100Kb0 (0/180)0 (0/180)0 (0/180)0 (0/180)51.95 (93/180)90 (162/180)500Kb0 (0/200)0 (0/200)3.5 (7/200)76.5 (153/200)100 (200/2)100 (200/200)1Mb0 (0/210)1.9 (4/209)64.13 (131/204)100 (210/210)100 (210/210)100 (210/210)2Mb4.76 (10/210)49.67 (100/202)97.58 (203/208)100 (210/210)100 (210/210)100 (210/210)5Mb73.99 (134/181)99.53 (201/202)100 (220/220)100 (220/220)100 (220/220)100 (220/220) Open table in a new tab