/, Collaboration/How collaborating with robots will speed up biomedical research with Oskari Vinko

How collaborating with robots will speed up biomedical research with Oskari Vinko



  • Big bottleneck in biomedical research is tedious routine work done by highly educated researchers
  • Current lab automation solutions are much behind what automation is possible in other fields like self driving cars
  • UniteLabs is working on a ‘plug-and-play’ solution for scientists who don’t want to spend weeks coding and setting up robots in their labs
  • They are already deploying their technology at Idorsia where their robots help with automating compound analytics
  • SILA is a communications standard that will help different lab devices talk to each other.
  • The lack of cross-disciplinary collaboration in life-sciences is a major bottleneck
  • Even scientists in high performing groups may work on niche projects because they rarely look outside their labs to engage in higher impact work and fruitful collaborations
  • One reason is the way the current incentive system works for scientists, which heavily favors first-author publications. This is often directly opposed to great collaborations.


Oskari Vinko is the cofounder of UniteLabs, a lab automation robotics startup based in Basel, Switzerland. He has a education background both in Mathematics and Engineering as well as Biotechnology and Synthetic Biology. His work experience includes being a founder and software engineer in various software companies.

Find Oskari Vinko LinkedIn

How did you come up with UniteLabs?

When I was doing research in synthetic biology at ETH Zurich, I noticed how I was often doing tedious manual work and trying to remember settings on devices. I would have to wait 10-40 minutes just to repeat the same process. I learned that everyone in research is doing this.

So I thought that there has to be another way.

While there are existing lab automation solutions out there, they are often difficult to use for people with no experience in programming and robotics.

In other words, the current lab automation is rather primitive compared to the capabilities we have with mobile phones and self-driving cars today.

So we started UniteLabs to bring the current, more advanced state of robotics & machine learning from other fields to speed up the life science research by freeing scientists of tedious routine work.

Who are the customers of UniteLabs?

Most people who work in labs may be quite uncomfortable if they have to scripting things in the lab. They are great scientists and biologists, but haven’t been trained in robotics and computer science. Hence they are unlikely to be able to use the existing solutions and really create high performing automated lab setups. UniteLabs’ goal is to make it really easy for less-technical scientists to create automated processes in their labs.

Customers include both people in pharma who use UniteLabs to set up drug discovery processes as well as people in academia who may be doing more fundamental research.

How does UniteLabs compare to cloud labs like Transcriptic? They seem to provide a turn-key solution with heavy automation.

The idea behind Transcriptic is that you can really just remote control a lab and send your requirements, samples and process to them and than they run the process for you and send you back the results. We think it is a great service and we look forward to this platform getting even more advanced than it currently is.

However, there are several reasons for when people still prefer running things in their own lab. Here are a couple

Transportation of delicate cells can be an issue, and you not want them to make a transatlantic trip before you can study them

Privacy is also a large concern. If the experiments you are running contain some of your most valuable IP, you may not want anyone else from seeing this before you’ve properly protected yourself.

Lastly, greater control over your experiments in real time is also a very important reason that people will still run many of their experiments in house. There is something about seeing the science happen in real time and your ability to tweak the process on the go which you won’t get in an automated cloud lab thousands of miles away.

What is a good case study describing UniteLabs?

One of our pilot projects is with a company is called Idorsia, an Actelion spin-out. We are working with their high-throughput screening unit, specifically in compound analysis.

This is one of the first steps before screening your compounds against a disease model is to verify the quality of their compounds. Especially if they have been stored for a while, in say a freezer.

Now we have built an automation system where the robotic system takes a plate with 96 chemical compounds that we wish to analyze. The plate contains a barcode, which gets tells the robots which process to run.

So the robot will take this plate and run the high pressure liquid chromatography where can separate the different components. Then the robot will take those components and, using another process, quantify them with mass spectrometry.

After a plate is completed, it may automatically go on to the next one and repeat the same process.

It is a relatively basic workflow, where we aim to prove the basic of the automation system we have built.

Different devices speak different languages. How do you bridge that gap?

What we need a plug and play system so that scientists can spend as little as time as possible on setting up an automated workflow.

Together with a few other groups, we are working on developing a new standard, upgrading the current SILA standard to SILA 2.0, a bit similar to bluetooth so that different devices can all be programmed and accessed by any other device.

Our thesis is that a universal experiment description language as well as interoperable data formats would be really useful. Then people could create workflows connecting different devices on through automation platforms. These could be shared on protocols.io, where people can benefit from each others research protocols.

What are other bottlenecks in research?

Negative Results aren’t shared

Pharma companies run tons of different research initiatives in parallel, even for just one drug. Many of those don’t work out and are scrapped. Similarly, lots of experiments in academia fail too.

When this happens, people often don’t feel the need to document the protocols and the failures, since they don’t have any immediate benefit from it. But others would greatly benefit from having access to that data and it could dramatically reduce unnecessary duplicate efforts.

Routine work needing lots of human input

Many processes are hard to automate, because they still rely heavily on human decisions. For example, in protein crystallization, which can appear a bit like black magic, some people have better intuition than others on which conditions could lead to the desired results. We are thinking of building a smart assistant which would allow a better collaboration between humans and robots. The robot would make it clear to the human what it can do and where it needs help. The human would only have to interact in the critical decision making parts, and most of the routine work could then be done by the robot. This would allow us to automate more of research process and save scientists valuable time.

Lacking cross disciplinary collaboration

Even really high performing teams can stay in a narrow box and sometimes spend vast resources on highly academic, ie niche pursuits. They think that all that matters is getting a paper published.

If they went out to talk to other researchers and industry first and asked a lot of questions, they would find out what the real problems are out there. This would likely lead to much more interesting collaborations. Scientists would work on stuff that has a real-world impact much faster than if they stay in their four walls and dream up projects to work on.

What is preventing more collaboration happening naturally in the life sciences?

Answer: The War for Reputation

For example a professor can have a requirement that a researcher publish 4 first author papers during a contract period. So collaborating on any project where you aren’t the first name author won’t help you fulfill your reputational requirements.

This is 100% disincentivizing collaboration, as per definition only one of the authors can have his name listed on the paper first.

We need to rethink those reputation incentive schemes, and viewing it through a game theory lense of payoffs with collaborate and defect mechanisms can be helpful.

Better solutions from the software world?

In software engineering, sites like stackoverflow and github have managed to create a new reputation system for software engineers. It is now standard practice for employers to ask for profiles from these sites along with more traditional résumés of jobs and accomplishments. On those websites, other engineers (read: peers) can assess the value of the contributions made.

Two core ways working with github and stackoverflow incentivize more collaboration in the software world

  • Units of contribution are smaller, compared to publishing a paper which can be years of work. On github, it can be a single line of code. On stackoverflow, it may be a great answer to a difficult machine learning problem.
  • The collaboration happens ‘post-publication’ of those contribution units, ie I can publish my work and others can benefit, ie we collaborate implicitly. We don’t even have to be in sync time wise or in the same place geographically. My reputation gets assessed by my peers who upvote my answers or star my source code.

On those platforms, helping others out is with great contributions the best way to get a lot of reputation. We are wondering how this might apply to research in the life science field.

We are seeing some promising data from people like Rachel Harding who has generated tons of collaboration through sharing her data and talking about her work online.

UniteLabs is working developing SILA 2 Browser, which will discover any lab devices who conform to the SILA 2 standard. The software will be open source, and hopefully many people can contribute to its development.

About Bio2040

There are so many challenges in biomedical research and drug discovery. We are a group of entrepreneurs and scientists who want to improve things. Our first measure is to educate ourselves and the community on what the biggest bottlenecks and their underlying reasons are. This leads us to discover exciting new opportunities. Bio2040 wants to be the leading place for or entrepreneurs & academics to meet, exchange ideas and launch new ventures.

How to stay connected or get involved

Subscribe to the Podcast on iTunes

Be in the know and sign up for the newsletter.

And you should follow Flavio on Twitter here.

By |2018-03-28T19:19:04+01:00March 28th, 2018|automation, Collaboration|0 Comments

About the Author:

Leave a Reply

%d bloggers like this: