Join MySLAS Social

Joshua Kangas: Building an Engine for a Start-up Business

While still a student, Joshua Kangas, Ph.D., made a life-changing decision. Instead of moving along his planned career path in education, he decided to partner with science and business experts to launch a company that would help enhance the efficiency of drug discovery efforts.


SLAS member and Journal of Biomolecular Screening (JBS) author Kangas was in the midst of his Ph.D. program at the Ray and Stephanie Lane Center for Computational Biology at Carnegie Mellon University (CMU) Pittsburgh, PA, when he and Professor Robert F. Murphy, Ph.D., had a big idea. The pair created and patented the Computational Research Engine (CoRE), a big data analytics technology designed to accelerate drug discovery and development by improving the efficiency of experimentation. As CoRE evolved, he and Murphy partnered with Scott Bodine and Geoffrey Hoare in 2012 to found Quantitative Medicine, Pittsburgh, PA, the company that would commercialize CoRE's technology.

While excited about the possibilities, the switch from academia to industry was not without a change in mindset for Kangas. "As a scientist in academia, I was constantly trying to explain my results in the clearest manner possible so that the results can be reproducible," he remarks. "As a scientist in industry, I still want to clearly describe our scientific process, but at the same time I need to choose my words carefully to protect our patented technology."

He had an opportunity to perfect his pitch when SLAS awarded Quantitative Medicine and seven other companies with exhibit space on SLAS Innovation AveNEW at SLAS2014 in San Diego, CA. This unique SLAS program fosters emerging talent by providing start-up companies with complimentary exhibit space, full conference registrations, travel and hotel accommodations. Interested entrepreneurs must apply for positions and their technical merit and commercial feasibility are considered by a panel of volunteer judges.

"We had great support from SLAS. The Society provided us with opportunities to meet with other scientists and demonstrate the value of what we are doing. The SLAS audience is interesting and a great resource for networking and collaborating on research solutions," says Kangas.

Introducing a New Research Tool

He is excited to demonstrate what big data analytics can do for research. CoRE blends machine and active learning algorithms to improve outcome predictions and selection in experimentation. "The goal is not to replace experimentation with computation, but to use computation to guide experimentation. Machine learning approaches can be used to develop models that make predictions about the outcome of experiments," Kangas explains. "These models can be improved by adding more data from new experiments. When there are a large number of potential compounds that can be tested such as in drug discovery, active learning approaches can play a crucial role in selecting only the most useful experiments. By pairing the two approaches, CoRE directs the experimentation process to explore very large experimental spaces that include the effects of potential drug compounds on numerous diverse targets." In its pilot tests, Quantitative Medicine reports a 40 to 90 percent reduction in required experimentation to reach client research objectives.

"Our clients typically have a tremendous amount of data for which they have already run analysis. When we run a demonstration, we have the client hide data from CoRE, as if those experiments had never been run," Kangas explains. "We iteratively reveal the hidden results as requested by CoRE as if it were actually directing experimentation. During this process, it typically finds the most important information much faster than the researchers did, showing prospective clients the value of CoRE."

Using high-throughput screening (HTS), scientists can run thousands of experiments and run machine learning algorithms on the resulting data, he says. "In developing CoRE, we turned that process on its end by conducting a smaller number of experiments and using machine and active learning algorithms on relatively small batches of the resulting data to selectively choose the next set of experiments," Kangas continues. "Initial commercial implementation will focus on discovery of small molecules. We originally developed a program to work in the HTS area because we imagined that logistically it would be easier to have robots that typically do HTS run those experiments, eliminating the need for a person to do the tedious work of managing many experimental results."

CoRE also has the capacity to leverage many different kinds of historical experimental results. Kangas says. "Many of the experiments in small molecule drug discovery and development take on a common form. One is simply measuring the changes in a system of varying complexity (from protein level effects in HTS to effects in humans in clinical trials) that results from adding a test compound. Because of this commonality across many experiments, CoRE can search for useful correlations across hundreds of millions of available historical experimental results." He explains that when scientists want to measure change in a complex system from a specific drug, they can use data from many different phases to help improve predictions because these experiments take an identical form.

At the Heart of CoRE

During his undergraduate years at Truman State University in Kirksville, MO, studying computer science, Kangas developed interests in not only artificial intelligence (AI) and machine learning, but also biology. The encounter with AI and machine learning was powerful, and he began to look for ways to use computer science approaches for solving biology problems. This led him to graduate studies in computational biology at Carnegie Mellon in 2008. He began thesis work under the supervision of Murphy, who introduced him to the need for active machine learning approaches to biomedical research problems. This was a major focus of discussions in the newly formed Ray and Stephanie Lane Center for Computational Biology, which Murphy directs. Those discussions, especially with Armaghan Naik, Ph.D., sparked the inspiration for the patented technology at the heart of CoRE.

"I looked at the problems the pharmaceutical industry faces as their research and development expenditures increase but the number of new drugs released remains about the same," he says. "I find the active learning process really exciting. It's incredible the things that you can get done by letting the program carefully choose the experiments to run next in light of what it has already learned."

The team at Quantitative Medicine has many incentives for expediting how the research process unfolds and progresses. CMU alumnus and company co-founder Bodine, who was diagnosed with Parkinson's Disease in 2007, developed a vision for what would ultimately become Quantitative Medicine as he underwent his first brain surgery in May 2011. Bodine articulated this vision with the surgeons who inserted a Deep Brain Stimulation (DBS) implant in his brain. Later that same year, Bodine discussed with Murphy some novel thoughts relating to the use, application and commercialization of biomedical data. By February of 2012, the discussions became focused on machine learning in drugs and diagnostics. According to Kangas, "I then met Scott and we all quickly determined there was a natural overlap between Scott's vision and what Dr. Murphy and I had discussed about the technology. If we can decrease R&D costs, we can increase the possibility of creating new drugs to treat relatively rare diseases where there might not be a financial incentive to do so as so few people are affected."

Bodine also brought his connections with the University of Florida Medical School and the Florida Innovation Hub at the University of Florida. The Hub, located in Gainesville, FL, is a science and technology business incubator that provides office space, laboratories and other resources. "We all agreed that teaming up would be effective because we could utilize resources from both the Innovation Hub in Florida and a Pittsburgh-based technology workspace, Revv Oakland," Kangas explains.

As the Quantitative Medicine team grew, so did Kangas' appreciation for what each team member brought to the equation. "The primary lesson I learned in forming the company is it's invaluable to have a team of people with diverse skills and interests," he says. "In my case, it has been very helpful to have people with significant business experience to help to shape the organization and business direction around the technical expertise in the company."

Kangas echoes what many other entrepreneurs say. A recent SLAS Electronic Laboratory Neighborhood article captured the stories of three SLAS members who managed successful business launches while struggling to wear all the hats. These members each expressed the importance of adding trustworthy, proficient partners or finding resources to bolster the areas in which they needed help. They cited struggles with everything from making a sales pitch, setting up customer service and basic accounting, to finding and motivating new company hires and having only 24 hours in the day. The members stressed the importance of listening – to clients, mentors and other entrepreneurs – to increase their education with every contact. They also credited involvement in SLAS and other business and professional organizations to build important working relationships.

Kangas realized the importance of all these aspects of business management, too, when noticing his able business partners deftly handle their areas of expertise. It was a win-win situation. "I would advise others to put together a team that is strong and experienced in starting companies in the sector in which you want to be. Leverage their abilities with your own. That is a more successful formula than trying to do it all yourself."

Something to Share

To escape his strenuous start-up business schedule, the father of two young sons enjoys getting out on the Allegheny River in his kayak. Pittsburgh sits at the confluence of three rivers: the Allegheny from the northeast and Monongahela River from the southeast join and form the Ohio River, making a wonderful watery environment for day trips.

"I like to stop on one of the islands in the river and have a nice peaceful lunch," Kangas says. "Kayaking is really relaxing and great exercise. It also gives me an opportunity to get a bit closer to nature as paddling a kayak can be done very quietly and in very shallow water. Occasionally this yields some fun fishing opportunities." And memorable experiences.

Once on a nighttime trip mid-summer, Kangas happened upon thousands of hatching mayfly nymphs emerging from the water. "At that point, they sit on top of the water for a short period to dry their wings. This makes them very vulnerable to fish attacks," he explains. "In that section of the river that night, I saw more species and diverse sizes of fish in one place than I had ever seen before in my life. I don't know if I would have been able to see all of that had I been in a noisy motor boat."

For the time being, Kangas will be spending most of his time in diverse scientific labs, explaining how CoRE can help direct their experimentation. In addition to showing scientists how to rethink research, Kangas also hopes to influence how data within large companies is shared. He hopes to enhance the "water cooler effect": the chance meeting of employees during breaks in which they discuss work and discover applicable ideas useful to their individual projects. "That's just luck," Kangas says. "If researchers place their data into a shared system and allow powerful machine learning algorithms to analyze their data, we can actively discover those relationships and increase opportunities to find shared solutions."

He also envisions researchers taking the active learning process a step further. "Rather than have only research groups within a single company sharing data, we could have people from multiple companies put data into our system," he explains. "Everyone is then leveraging what's in the system. It would really improve the efficiency of R&D."

Kangas plans to continue making connections among SLAS's diverse membership and to share both his ideas about shared data and his start-up business knowledge with others. His spot on SLAS Innovation AveNEW also came with an invitation to present his capabilities during the popular Late Night with LRIG Rapid Fire Innovation session. Kangas also represented his published work at a JBS Meet and Greet in the SLAS Member Center at SLAS2014. These activities gave Kangas ample time to share CoRE's capabilities.

To expand opportunities to connect with those interested in launching a company, Kangas joined the SLAS Student and Early Career Professionals Advisory Committee, which ensures that the Society's programs for students and early career professionals are aligned with the SLAS mission and goals. "There are people in this community from academia and others from large corporations," he says. "I represent a somewhat different perspective – a start-up company. I hope to share what I have learned with others who may be interested in taking on such a venture."

September 15, 2014