Emedgene: AI for Genomics

By Tarmo Virki | genes | February 15, 2018

This article is published in CoFounder No 11 in February 2018.

Emedgene is a big data startup helping human geneticists diagnose complex genetic conditions within minutes at the point of care. The team of a few dozen uses artificial intelligence and novel scientific methodologies to automatically identify unique mutations, helping make data interpretation process faster and easier.

The technical co-founders, Niv Mizrahi and Ofir Farchy, had worked side-by-side for years at an AdTech company before they started to look for ways out. “We were there for almost seven years and decided that we wanted to build something of our own,” Farchy told CoFounder in an interview.

They first took a stab at the legal tech field. “It took us 4-5 months to understand that the market was not happening yet. We thought it was a brilliant product, but the market thought differently,” Farchy remembered.

At the same time, Einat Metzer was starting to look for technological founders to build Emedgene. She had met Shay Tzur, a brilliant geneticist, and they understood that there must be a technological solution available for the work Tzur was doing by hand. He was going through thousands of lines of Excel tables to spot mutations that could be the reasoning for some disease. It took him at best a few days, but often weeks, to carry out each analysis.

The four met each other for the first time in 2014, when the legal tech duo faced the big decision: to pivot or start something new. “That’s when Niv first met Einat. It was like love at first sight: he decided it was an amazing idea. We clicked together really well. It was the first time I had met Shay,” Farchy remembered.

“We decided that instead of making lawyers lives better we want to make the world better … by getting access for everyone to get genetic tests done if they need them,” he said.

Four founders is probably a lot for many venture capitalists, but a good division of labour helps to make things work for Emedgene team.

Einat Metzer, coming from the VC world herself, was clearly the business person in the team from the start. Shay Tzur brings in the genetics know-how. “No one can rival Shay’s experience in being a geneticist. We tried to learn everything he knows from him, we tried to model everything he knows and all of his intuitions, we tried to model it into machine learning models,” Farchy said. The main task was splitting the roles of the technical founders, but they had known each other for years, so it was easier than it could have been. Mizrahi took the tech manager role, while Farchy became the CTO, leading the research.

The cost of DNA sequencing has dropped and use of it has started to increase. What Emedgene is addressing is the growing pain-point: analysing the increasing amount of data.

“We’re starting to see more and more governments budgeting for this, and hospitals starting to provide more genetic tests,” Farchy said.

The work of machine sequencing usually results in a digital file of tens of gigabytes. “What used to happen is that people would sit at the end of the process in front of an Excel file, with tens of thousands of lines, and research each and every line to understand if that was the genetic mutation they were looking for,” Farchy described the work challenges of geneticists. “We’re trying to do the entire process, including working through the Excel file – we try to do that automatically. What we have managed to do is reduce the time from a couple of days or weeks to a matter of minutes. By doing that we are enabling genetic labs to keep their current staff and scale up, making it more accessible for everybody.”

The company has just swapped to using Microsoft’s Azure cloud, supported by the firm’s startup programmes. “We did three separate days of hackathons with their developers. They gave us the most experienced cloud architects to sit with us in the same room. That was extremely helpful. You feel like you have a personal connection with the people that actually write the code for the cloud. That’s a first for me,” Farchy said.

For anyone thinking about their web connections and files in tens of gigabytes, Farchy has a calming message: most of the sequencing files go direct to manufacturers’ special clouds, and data connections between the clouds are typically very fast. “Luckily they have bandwidth that I just wish I had in my house. If I had to do this at home it would probably take me a couple of days to transfer a single file,” he laughed.