NEW YORK (GenomeWeb) – Seattle-based Spiral Genetics has shifted the focus of its business from building solutions based on existing open source software for genomic data analysis, to developing a suite of products based on a proprietary method of detecting large structural changes in the genome.
Spiral has since launched two products — Anchored Assembly and Onco Assembly — based on the new method and plans to launch additional products for other genomic domains in the near future, Spiral CEO Alindrina Mangubat told BioInform this week.
Spiral has also discontinued development on the platform it initially offered to the market. The company used to sell cloud- and cluster-based software that used common open source bioinformatics tools such as BWA and the Genome Analysis Toolkit for early stage genomic data analysis tasks such as sequence alignment, consensus calling, and variant detection.
Spiral planned to pursue partnerships with bioinformatics vendors that were already developing downstream analyses tools for processes, such as variant filtering and annotation, and to couple its platform with theirs to offer more complete solutions to the market. It signed one such agreement with Omicia allowing both companies to jointly offer a complete solution for analyzing raw sequence through to identifying clinically relevant genomic variants. Spiral raised $3 million in series A financing last year from Draper Fisher Jurvetson to expand its bioinformatics engineering team, improve sales, and develop new features for its platform.
Things changed for Spiral shortly after its fundraiser. As Mangubat explained to BioInform, Spiral's development team had been working on a method of accurately detecting large structural variants using short-read sequencing data. They did so to provide an alternative to current methods, which, she explained, are good at finding small changes in the genome such as single basepair mutations. But they are less accurate when it comes to locating variants of much larger sizes — for example, a mutation that is several thousand basepairs long — and they produce results that have high false positive rates.
One solution is to use longer-read sequencing technologies to generate the data, but Spiral believes it has found a way to accurately detect these mutations in standard short-read data and to do so in a manner that scales to accommodate increasing quantities of data.
According to Mangubat, the secret to the method's success is its approach to read alignment, which is essentially the inverse of a mechanism used in many existing reference-based assembly methods. These methods first align all sequences that map to the reference and then perform local de novo assembly for reads that vary significantly from reference.
The problem with this approach, however, is that "you are biasing everything that you are doing against the reference genome," Mangubat told BioInform. So "if you have a 3,000-basepair insertion and your reads are only 100 base pairs long, then there's going to be whole swaths of things that don't map anywhere to the reference genome and that information will essentially be lost." An alternative approach is to try to assemble all of the reads de novo but that's a computationally expensive process.
Spiral's approach, on the other hand, compares all the unmapped reads to the reference, but as a first step defines only the reads that are not perfect matches to the genome initially and assembles these de novo, Mangubat explained. After that, the software maps the reads that already match the reference. "By doing that we can assemble huge variations and we are not biased against the reference because we did not align everything first and then try and go backwards," she said. In fact, internal estimates provided by the company on its website claim that the method detects about 30 percent more structural variants than the current gold standard with a 3 percent false discovery rate.
One set of comparison tests that the company has done was performed in collaboration with researchers at Baylor College of Medicine. Mangubat said that the partners compared Spiral's software with some existing variant detection software as well as some chemistry technologies. The results of that particular study have been submitted for publication, so Mangubat could not go into specifics about the findings. However, she said that the results show that her company's method did better at finding structural variants compared to the other approaches tested as part of the study.
The company has also published details of comparisons between its Anchored Assembly solution — which encapsulates the company's core methodology — and two variant detection pipelines, BWA+GATK and BWA+Pindel. The results show that the Anchored Assembly software has greater sensitivity across a range of variants of different sizes and lower false discovery rates than the two other solutions. Details of that comparison are available in a white paper on Spiral's site.
Spiral launched Anchored Assembly last spring. The product is used to detect insertions, deletions, transpositions, repeats, as well as single- and multi-nucleotide variants in sequence reads from Illumina instruments. The software is well-suited for Illumina X Ten, HiSeq, or MiSeq reads, with at least 20x coverage per chromosome set, according to the company.
More recently, the company released a specialized version of the Anchored Assembly software for cancer data analysis. That product is called Onco Assembly and it is tailored to detect and home in on tumor-specific variation. The software works well with data from solid tumors but it could potentially work on data from hematological cancers as well, Mangubat said.
Both products are available either as cloud-based solutions or installed locally on in-house compute clusters. Pricing for Anchored Assembly starts at around $700 to $800 per whole human genome — this price point assumes 30X to 35X coverage and scales as the depth of coverage increases. There is a volume-based pricing option that can bring the cost per whole genome down to about $300 to $400 depending on the number of samples, the amount of customer support needed, and so on. Spiral also offers flexibly priced site licenses for Anchored Assembly — costs here depend on factors such as the volume and capacity of the sequencers as well as the kind of data being produced.
This level of pricing is most cost-effective for larger research facilities and centers that have and use a lot of sequencing capacity in house, and this is primarily the segment of the market that Spiral is targeting with its product, Mangubat said, although the company is also open to working with smaller research groups. Pricing for Onco Assembly is also flexible and is determined by factors such as the nature of the project, number of samples, the purity of the tumor sample, and sequencing coverage. The company's solutions are being used at institutions like the University of California by researchers at the Davis and Irvine campuses, as well as at the University of Melbourne and the University of Montreal.
Spiral is exploring other potential applications of its technology and mulling products for things like trio analysis and agrigenomics studies to, for example, differentiate between affected and un-affected cultivars, Mangubat said. Meanwhile, the company is also working on new technology that will enable the comparison of thousands of genetic samples at a time in studies aimed at identifying variants that are associated with particular phenotypes, she said.