Wednesday, February 14, 2018

AGBT: Twist Biosciences Launches Sequence Capture Product

Twist Biosciences today launched a new product into the sequence capture space.  CEO Emily Leproust was presenting to the Gold Sponsor workshop as I started writing this, but she also sat down with me yesterday to preview the new offering for targeted sequencing.
Targeted sequencing, particularly exome sequencing, was once seen as a short-lived transitional technology until the cost of whole genome sequencing was brought down substantially.  We now have $1K human genomes, but exomes remain a popular option for obtaining data on more samples at a lower cost than possible with whole genome sequencing. 

Twist sees opportunity in this market by providing a product delivering greater coverage uniformity, enabling exomes to be acquired with much less overall sequencing.  This lowers costs and enables more samples to be run.  In a comparison against what Twist believes to be the best competing product, an exome required only 3Gb of data to achieve 20X coverage over 93% of the targeted regions, 70% less than the 5.1Gb required by the competitor.  On another example, Twist can pack 267 samples onto a NextSeq run versus 187 for the competitor.

How is this achieved?  Twist has developed their semiconductor-based oligo synthesis in a manner that delivers very high synthesis uniformity.  These initial oligo pools (1 fmol/oligo) be amplified to form the probe pools (1pmol/oligo), a place where Twist has also focused on developing procedures which deliver great uniformity.  Twist also has extensive experience with oligo design and performance in their gene synthesis business, and has used this to ensure that capture efficiency is minimally dependent on G+C content.  Twist also analyzes each pool by high-throughput sequencing and discards pools that do not meet their quality standards.  Twist also believes their double-stranded DNA probes outperform single stranded probes.

In addition to marketing standard panels, Twist will offer a custom panel option.  A key part of this offering is a rapid design-build-test-analyze cycle, with the design and build at Twist in 2 weeks.  Performance with the initial design can be assessed using the FASTQ files and the capture design further tuned for maximal sensitivity and uniformity and for reduced off-target capture.  In a test, the initial design had 64% off-target capture which was reduced to 21% and finally to 7.6%.  Twist believes that maximal performance can be obtained with only 2-3 rounds of optimization, potentially taking less than 2 months to obtain an optimal design. 

Early Adopter: Shawn Levy, HudsonAlpha

Shawn Levy from HudsonAlpha gave an overview at the workshop on using the Twist product for capturing from FFPE samples. 50 nanograms of DNA is fragmented, end repaired and oligo tailed in a single tube reaction.  After purification, it is ligated and then purified again and amplified.  In the experiment described by Levy, two pools of eight samples were created.  One pool contained just tissue culture and normal DNA, but the second pool contained four FFPE samples and four "pristine" samples.  Capture was performed on pools. These were sequenced on a NovaSeq to generate about 500X coverage.  Even though pool 2 contained lower quality samples, relatively uniform capture was still achieved.  

Clearly 500X is not standard coverage, so the dataset was downsampled to 4, 6, 9 or 15 of data.  Even then, the FFPE samples retained acceptable uniformity.  Levy commented that some discordant calls in the dataset may have been due to index switching, which he believes will be significantly reduced in future indexing primer schemes for the Twist kits.  Even in downsampled FFPE data, over 99.8% of variant calls were concordant with the same samples run with the Roche NimbleGen capture product. 

Levy imagines that the future would see 96 samples being converted to 12 hybridization pools which in turn would be superpooled to form 6 sequencing superpools and 4 of these superpools could go on a NovaSeq S4 flowcell to generate 10Gb per sample for 100x coverage exomes.  

Given the problem with index hopping, Levy is a proponent of using unique i5 and i7 indexes in experiments as well as carefully assessing which index combinations perform well.

Emmanuel Martin, IntegraGen

Early Adopter: Mike Wiley, USAMRIID

Twist has already been developing custom kits.  In conjunction with the U.S. Army Medical Institute of Infectious Disease (USAMRIID), Twist has designed a capture kit for detecting viruses in patient samples, which covers all known viral pathogens.  This can then be used with DNA or with RNA-Seq libraries prepared with Ilumina's TruSeq RNA kits. 600K probes covering 1000 RefSeq viral segments are in the design.  

600K probes, 1000 RefSeq viral segments -- DNA, RNA, ss, ds, segmented


No comments: