NEW YORK (GenomeWeb) – The developers of a machine learning-driven genotyping algorithm for analysis of high resolution melt (HRM) curves have launched a startup to commercialize the technology.
San Diego-based HeatSeq was founded in 2013 by Samuel Yang, now at Stanford University, and Stephanie Fraley, now at the University of California, San Diego, and is looking to create a PCR back-end platform that uses melt curve analysis to fingerprint DNA sequences and catalog them.
The startup has exclusively licensed patent-pending intellectual property from Johns Hopkins University, Fraley told GenomeWeb, including the algorithm, the method of doing HRM, and a key reagent used in the process. Both Yang and Fraley were at Johns Hopkins when they developed the technology.
The initial application they're pursuing for the technology is identification of bacteria in infectious disease. The algorithm is able to identify and quantify the presence of multiple bacteria species in a blood infection, Fraley said, by turning DNA melt curves obtained during PCR and HRM into a 300-point data set and comparing those data to HeatSeq's proprietary database of known bacterial DNA melt curves. Different DNA sequences melt at different temperatures, based on the specific sequence of DNA.
To ensure detection of bacterial DNA, the method uses primers taken from universally conserved regions in the bacterial 16s rRNA gene that bracket highly variable regions. The variable regions allow the technology to identify different bacterial species; in some cases, the melt curves can detect single nucleotide differences between the variable regions.
Another important part of the technology is the reagent HeatSeq has licensed from Johns Hopkins. "The chemistry promotes very specific amplification and also reliable melting curves," Fraley said. "The algorithm and reagent rely on each other and allow the technology to do as good of a job as it does."
She further explained that melt curves are very sensitive to changes in reaction chemistry, salt content, or pipetting errors, for instance, making it extremely difficult to get consistent results between different lab users. But HeatSeq's reagent allows the algorithm to normalize results "from user to user, machine to machine, and day to day," Fraley said. "That's part of why our technology is advancing melt curve technology to where it can be a primary identification tool instead of a secondary check."
HeatSeq is about halfway done with its bacterial pathogen database, which contains the approximately 200 bacterial pathogens that are responsible for almost all infections. It is also working on a pan-fungal database, but the technology is not limited to detecting pathogenic species present in sepsis. Fraley said it could be a platform for detecting multiple nucleic acids present in a sample, such as in characterizing tumor suppressor genes and let-7 miRNA for oncology.
HeatSeq is looking to partner with a PCR front end, with the ideal partner being either traditional or digital qPCR instrument that is fully automated and potentially even portable, Fraley said. "Our back end is completely automated so it would benefit from companies who have that figured out on the equipment side. There's a lot of value in partnering with someone who has the front end completely automated as well," she said. Though Fraley declined to name them, she said several small and large PCR instrument makers have already reached out to HeatSeq.
The firm has already received grants from the National Science Foundation I-Corps program and the Stanford Predictives and Diagnostics Accelerator, though Fraley declined to disclose the grant amounts. "It's not a huge amount," she said. "We're still very much in the fundraising stage."
With no lab space and no employees, HeatSeq is still very early stage. However Fraley said the firm has lined up options for space in wet lab biotech incubators in the San Diego area in anticipation of funding, and plans to hire software developers, scientists, and technicians to work on the algorithm and databases.