Menu
October 19, 2016  |  General

Identifying Structural Variants in NA12878 from Low-Fold Coverage Sequencing on the PacBio Sequel System

Recent de novo assemblies of individual human genomes have uncovered thousands of structural variants, many of which are accessible only with PacBio long reads [1-3].

Personal Genome PacBio Coverage Deletions ≥50 bp Insertions ≥50 bp
CHM1 [1] 41-fold 6,111 9,638
HX1 [2] 103-fold 9,891 10,284
AK1 [3] 101-fold 7,358 10,077

A similar increase in structural variant sensitivity relative to short-read methods has been demonstrated with low-fold coverage PacBio sequencing interpreted against the reference genome [4].  To demonstrate and evaluate the low-fold coverage approach on the PacBio Sequel System, we generated approximately 10-fold coverage of the well-studied human sample NA12878.

Methods
Purified DNA for NA12878 was obtained from Coriell, sheared to an average size of 25 kb, converted to SMRTbell templates, and size selected to 15 kb on the BluePippin system (Sage Science). The resulting library was loaded on 10 SMRT Cells. Each SMRT Cell was run for 6 hours on the Sequel System with chemistry v1.2 (an older chemistry than was used for recently released Arabidopsis data, which uses the newer chemistry v1.2.1 and has a yield of about 5 Gb per SMRT Cell and read length N50 of 16.4kb).  In total, the runs generated 32.8 Gb of data contained in 3.4 million reads with half of the bases in reads longer than 11.8 kb.
Sequencing Metrics

SMRT Cells 10
Run Time 60 hrs
Number of Bases 32.8 Gb
Number of Reads 3.4 M
Read Length N50 11,823 bp

Reads were mapped to the GRCh37 human reference genome with NGM-LR [5], and structural variants were called with PBHoney [6].  A total of 7,386 deletions and 7,445 insertions of at least 50 bp were identified and comprise the “10-fold SV call set.”

Visualizing Structural Variants
Ongoing improvements to the IGV browser [7] (available now in the development version) improve visualization for PacBio reads and structural variants. With these updates, IGV provides a clear representation of deletions, insertions, and trinucleotide repeats, and shows how long reads span structural variants.
Heterozygous 315 bp deletion at chrX:116,454,160-116,454,859
chrx_116454160_116454859

Homozygous 328 bp insertion at chr10:92,213,800-92,216,245
chr10_92213800_92216245

FMR1 trinucleotide repeat small expansion at chrX:146,993,200-146,993,950
chrx_146993200_146993950

Evaluation of 10-fold Call Set
To quantify sensitivity, the 10-fold SV call set was compared to a merged NA12878 “truth” set from the 1000 Genomes Project [8] and Genome in a Bottle [9].

Set Platform Deletions ≥50 bp Insertions ≥50 bp
truth: 1000 Genomes + GIAB [8,9] Illumina 3,021 1,090
10-fold SV call set PacBio Sequel 7,386 7,445

The 10-fold SV call set recalls 86% of truth set deletions and 81% of insertions.  Moreover, it includes thousands of deletions and insertions that are not in the truth sets, most of which are directly confirmed by a FALCON-Unzip de novo assembly from 60-fold PacBio RS II coverage.

In summary, this 10-fold SV call set demonstrates that low-fold coverage sequencing on the PacBio Sequel System is an affordable, effective approach for identifying structural variants and provides much improved sensitivity compared to short-read approaches.  We are excited to see how this approach will be extended and applied to study genetic variation in disease cohorts, in human populations, and in other organisms.

figure-2
Data Availability
To illustrate the low-fold coverage structural variant calling workflow, the NA12878 Sequel data is available for analysis on DNAnexus.

[1] Chaisson MJ, et al. (2015). Nature, 517(7536):608-11.
[2] Shi L, et al. (2016). Nat Commun, 7:12065.
[3] Seo JS, et al. (2016). Nature, 538(7624):243-7.
[4] English AC, et al. (2014) BMC Bioinformatics, 15:180.
[5] https://github.com/philres/nextgenmap-lr
[6] English AC, et al. (2015). BMC Genomics, 16:286.
[7] Robinson JT, et al. (2011). Nat Biotechnol, 29(1):24-6.
[8] Parikh H, et al. (2016). BMC Genomics, 17:64.
[9] Sudmant PH, et al. (2015). Nature, 526(7571):75-81.

Talk with an expert

If you have a question, need to check the status of an order, or are interested in purchasing an instrument, we're here to help.