Startup Fund Company Acquired
On December 17, 2019 Parabricks, a portfolio company of the Michigan Biomedical Venture Fund (MBVF) was acquired by Nvidia Corp and became a part of Nvidia Healthcare. This acquisition of Parabricks was the MBVF’s first exit and occurred less than two years after our first investment in the company.
In early 2016, two recent University of Michigan PhD graduates Mehrzad Samadi and Ankit Sethia, along with their advisor, Professor Scott Mahlke, were looking for a problem to solve and to put their graduate work to some practical use. The computer science team had a number of years of experience writing parallel code for Graphical Processor Units (GPUs). Instead of using computers with CPUs that execute lines of computer code sequentially, the newer GPUs enable the execution of code in parallel. This allows multiple streams of data to be processed in parallel leading to speedup.
Eschewing lucrative jobs in the tech industry, the two graduates formally launched their PhD project as a startup company they named Parabricks.
First, they attempted to solve problems in the automotive sector and received early translational funds from the state-funded MTRAC Transportation program, mentorship through the Office of Tech Transfer, and training through the National I-Corps program. They quickly realized that the need was not as great as they had initially imagined. They knew they had great technology but were looking for an application where they could apply these techniques.
The problem that desperately needed solving was present in the nucleus of every cell in the human body. The coiled strands of DNA encode proteins and are the basis of the functions of life. The human genome is encoded in the DNA and consists of ~3 billion base pairs arranged in a double helix, a structure famously deciphered by Watson and Crick in 1953. The cost to sequence the extremely large human genome has dropped from $2.3B for the first one sequenced in 2003 to less than $699 today. The resulting drop in costs has resulted in an increase in the number of genomes being sequenced. By some estimates, the data generated by the 3 billion DNA base pairs would generate 20 ExaBytes – more than Twitter, YouTube, and scientific astronomy data.
While costs have dropped during the past decade, the process of obtaining a DNA sequence is still long and laborious because it cannot be obtained end-to-end in a single step. The DNA is first broken down into smaller fragments, then a sequencing machine is used to detect the order of the bases. Then billions of these fragments are then reassembled to construct the single strand of the whole genome. This process typically takes more than a few days to complete. As the sequencing becomes more ubiquitous, more data is being generated. Data analysis is a major bottleneck from a time and cost perspective. The Parabricks team had found a real need for their complementary skills in parallel coding.
By late 2017, the team had already made great progress in reducing the processing time, but they still needed some time and money to continue to refine and improve their solution.
While the market need for their solution in genomics was great and they had superior technology, the two young entrepreneurs found putting together funds for their fledgling startup a much harder problem to solve than putting genomic sequences together.
“Luckily for them, the timing of their launch was fortuitous. The University of Michigan had just launched a new early-stage seed fund in 2016 called the Michigan Biomedical Venture Fund (MBVF) to invest in companies with U-M licensed IP,” said Program Manager Hirak Parikh, PhD. “We were looking for promising startups that needed help bridging the proverbial ‘valley of death.’
“We found a great fit in the Parabricks team and were thrilled by their enthusiasm to tackle a problem that a number of large companies had tried to address but failed in the past.”
With funding secured from MBVF, the Parabricks team got to work. In addition to MBVF, the team also got early support from Invest Michigan and Invest Detroit. Eventually, Parabricks was able to offer 48-times faster processing of whole-genome sequencing (WGS), lower compute costs by 75%, and offer quality that met the gold standard for next-generation sequencing (NGS). Adding other proprietary speed-up tricks resulted in the team being able to reduce the time to sequence a whole genome from a week to under an hour: faster and cheaper than any comparable solution.
This was followed by an early partnership with Nvidia and subsequent partner relationships were also built with leading H/W manufacturers such as IBM, HPE, Inspur, Dell and Intel. In addition, Parabricks announced partnerships with provider DNA Nexus, primary sequencing company Veritas, and major cloud computing services such as Microsoft, Amazon, and Google’s Deep Variant Genomic Analysis tool. Parabricks was now working with hardware manufacturers and genomic sequencing platforms.
These new partnerships and growth did not escape the notice of their earliest collaborator and partner Nvidia. In late November, Nvidia decided to acquire the company.
“Next generation sequencing is in high demand with national sequencing programs clinical genomics and is driving personalized medicine,” explained Kimberly Powell, NVIDIA Vice President of Healthcare. “We are excited to have the Parabricks team join NVIDIA and double down on providing fast and accurate genome analysis tools to take genomics to the next level.”
As part of Nvidia, Parabricks will continue to advance technologies using GPUs to accelerate the analysis of genomic sequences. Their plan is to introduce products for faster single-cell sequencing and RNA in 2020.
“We plan to provide accelerated tools for major established genomic analyses, continue supporting all major versions of these packages in the long term, and include invaluable enterprise grade support to many major institutions,” said Mehrzad Samadi, Parabricks co-founder and CEO.
As costs of primary sequencing continue to drop, the demand for WGS for humans, animals, and crops is going to increase. By reducing compute costs Parabricks saves companies expensive cloud storage costs or reduces CPU cycles on local hardware that are needed to process the voluminous data. The current product serves the NGS market by speeding up the secondary analysis – alignment, pre-processing and variant-calling. Parallel computing algorithms can easily be adapted for other tasks such as RNA sequencing, protein folding or financial analysis as possible extensions of the core technology.
“We are extremely thrilled with the outcome and are really grateful to the Michigan Biomedical Venture Fund who provided two rounds of capital to the company at the critical early juncture in our history, enabling us to focus on company building and not worry about raising capital,” said Ankit Sethia, Parabricks co-founder and CTO. “We hope that the fund continues to support new entrepreneurs who are bringing new disruptive and transformative U-M technologies to market.”
MBVF has nine companies in its portfolio and invests in the domain of therapeutics, devices, diagnostics, or health IT. The Fund is a collaborative effort between the U-M College of Engineering’s Center for Entrepreneurship and U-M Medical School’s Fast Forward Medical Innovation program. Seeded with a $3 million gift from the Monroe-Brown Foundation, MBVF awards $50,000 – $300,000 to biomedical startup companies.