Automated genomic interpretation of sub-chromosomal fetal copy number variation by sequencing circulating cell-free DNA from maternal plasma

Issue link:

Contents of this Issue


Page 0 of 0

RESULTS RESULTS Figure 6. Discrepancy Resolution • 12 Samples were chosen for re-evaluation by a subset of the initial curation teams. • The focus was on samples called VOUS by the working group and pathogenic/Likely pathogenic (score 1 or 2) by Mantis. • 3 of the 11 VOUS calls were revised to pathogenic or likely pathogenic. • 2 of the 11 were upgraded to likely pathogenic by some team members but not by others. • One sample with an initial call of pathogenic was downgraded to VOUS. Figure 2. Microdeletion and microduplication detection • Aligned Reads are partitioned into non-overlapping 50Kb bins. • Bin filtering is applied to remove genomic regions with excess noise. • Bin counts undergo pre-processing to remove GC artifacts and other noise factors. • Circular Binary Segmentation (CBS) is used to detect potential regions of amplification and deletion. • Out of ~4500 patient samples 91 putative CNVs were detected and used for the current study design. Figure 3. Manual curation of CNVs by in-house experts • Working groups of lab directors and genetic counselors provide manual curation of the putative CNVs (n=91). • Annotation involves use of publicly available databases (ISCA, Decipher, OMIM) and visualization tools. • Consensus opinions are generated among the expert panels based on CNV size, overlap and frequency of database occurrence. • Scoring of results into Pathogenic, Likely Pathogenic and Variants of Unknown Significance (VOUS). Figure 4. Automated annotation and clinical interpretation • Mantis™ scoring system utilizes prior knowledge of previously observed CNVs and information regarding the functional impact on known disease genes. • Suspect CNVs are scored with a numeric system corresponding to pathogenicity rank. Figure 5. Comparative analysis of manual and automated results • 56/59 pathogenic or likely pathogenic samples were given a score of 1 or 2 by the automated method. This result suggests high sensitivity. • 17/32 VOUS calls were treated as pathogenic or likely pathogenic by the Mantis™ system (score of 1 or 2). • 15/32 VOUS calls were given a score of 3 or 4 by the Mantis™ system. • Discrepancy resolution focused on a subset of the discordant calls. Cypher Genomics Mantris™ 31-41502R1.0_0315 OvERviEw Genome sequencing-based noninvasive prenatal testing (NIPT) has been largely focused on whole chromosome events, such as chromosome 13, 18, 21 and sex chromosome aneuploidies. Collectively, whole chromosomal aneuploidies account for approximately 30% of all live births with a chromosomal abnormality. Sub-chromosomal structural variations, such as copy number variation (CNVs), account for the remaining 70% of live births with chromosomal abnormalities, but are more difficult to detect and classify. The physical boundaries of CNVs causing similar diseases can vary greatly because of the de novo nature of the events. Moreover, because of the relative rarity of each individual event, the critical region of many known pathogenic CNVs is not well defined. In order to evolve NIPT to a more comprehensive test, an automated method for the classification of pathogenic events is required for routine reporting to clinicians. In this study, we evaluated an automated CNV interpretation solution, Mantis™, by comparing classifications for 91 CNVs to a consensus opinion of experts. Mantis evaluates variants for likely functional impact using in silico prediction algorithms in conjunction with known annotated variants from a broad range of curated databases. The test CNVs were comprised of both duplications and deletions, spanning a range of ~1-77 megabases in size. The initial expert review resulted in 59 of 91 CNVs classified as pathogenic or potentially pathogenic. The remaining 32 were considered variants of unknown significance (VOUS). A comparison of expert opinion to classifications produced by Mantis suggest high concordance of the automated interpretation with expert review, providing a scalable and automated solution for reporting clinically relevant CNVs. Post-hoc discrepancy analysis resulted in VOUS resolution and highlighted the potential for increased diagnostic yield of NIPT via automated classification. RESULTS REfEREnCES 1. Jensen TJ 1 , Zwiefelhofer T, Tim RC, et al. High-throughput massively parallel sequencing for fetal aneuploidy detection from maternal plasma. PLoS One. 2013;8(3):e57381. doi: 10.1371/journal.pone.0057381. Epub 2013 Mar 6. 2. Zhao C, Deciu C, Ehrich M, et al. Detection of fetal subchromosomal abnormalities by sequencing circulating cell-free DNA from maternal plasma. PLoSone. In press. Figure 1. Noninvasive prenatal detection and annotation of sub-chromosomal CNVs Automated genomic interpretation of sub-chromosomal fetal copy number variation by sequencing circulating cell-free DnA from maternal plasma Penn Whitley 1 , John Tynan 1 , Adam Simpson 2 , Tim Burcham 1 , Ali Torkamani 2 , Ashley Van Zeeland 2 , Mathias Ehrich 1 , Ron M McCullough 1 , Juan-Sebastian Saldivar 1 , Tom Monroe 1 , Paul Oeth 1, and Dirk van den Boom 1 1 Sequenom Laboratories™, San Diego, CA; 2 Cypher Genomics, Inc., San Diego, CA COnCLUSiOn • The manual curation of Copy Number Variation in a NIPT setting is time consuming (~0.5-4hr/CNV/curator), subjective and potentially subject to error. • Data provenance and standards are also difficult to maintain in an environment with manual curation and interpretation. • The Cypher Genomics Mantis™ System provided highly concordant results with those from the working group when the initial call was pathogenic or likely pathogenic. • Manual VOUS calls were less concordant with the automated results. This finding is likely reflective of the different data sources and scoring systems used by the methods. • Discrepancy analysis resulted in several scoring reversals when the working groups re-evaluated the CNV calls. The reasons for these reversals may have been a result of updated data sources or more consistent scoring methodology used during the re-evaluation. • These results suggest that an automated interpretation and curation system can provide an excellent 'first line of defense' in a clinical setting. • The combination of the standardized output and interpretation generated by the Mantis™ system, along with the experience and judgment of a lab director, should provide excellent results to clinicians. iniTiAL MAnUAL EvALUATiOn MAnTiS™ SCORE UPDATED MAnUAL EvALUATiOn Pathogenic 3 VOUS VOUS 1 Pathogenic VOUS 2 Likely Pathogenic VOUS 2 Likely Pathogenic VOUS 2 VOUS VOUS 2 VOUS VOUS 2 VOUS VOUS 2 VOUS VOUS 2 VOUS VOUS 2 VOUS VOUS 2 *VOUS/Likely Pathogenic VOUS 2 *VOUS/Likely Pathogenic MAnUAL SCORinG SYSTEM MAnTiS™ SCORE COnCORDAnCE Likely Pathogenic 1 Sensitivity Pathogenic 2 VOUS 3 Specificity Likely Benign 4 Benign 5 MAnTiS™ SCORES MAnUAL SCORinG SYSTEM 1 2 3 4 5 TOTAL Likely Pathogenic 11 12 1 1 0 25 Pathogenic 30 3 1 0 0 34 VOUS 1 16 8 7 0 32 Likely Benign 0 0 0 0 0 0 Benign 0 0 0 0 0 0 Total 42 31 10 8 0 91 • cfDNA isolation from plasma of pregnant mothers • Single-end Illumina HiSeq • Average coverage ~0.2X • 4500 patient samples evaluated • 50Kb binning • GC correction and normali- zation • CNV detection based on read depth • Manual curation of detected CNVs by working groups of genetic coun- selors and lab directors • Scoring of curated results according to pathogenicity • Consensus opinions generated • Blinding of CNV events • Automated interpretation via Cypher Genomics Mantis™ pipeline • Reporting of results • Comparison of manual and automated calls • Discrepancy resolution Low Coverage WGS CNV Detection Expert Manual Curation and Interpretation Automated Interpretation via Mantis™ Post –Hoc Analysis

Articles in this issue

view archives of sequenom - Automated genomic interpretation of sub-chromosomal fetal copy number variation by sequencing circulating cell-free DNA from maternal plasma