A guest post by Meredith Drosback, Becky Hazen, & Rick Weiss The COVID-19 pandemic has vastly increased journalists’ demand for access to…
If you are a scientist, in particular a biologist, you might be very familiar with the frustration that comes from trying to reproduce an experiment described in a paper authored by someone else. No matter the journal where the research is published, reproducibility in science is one of the practitioner’s biggest concerns. A recent survey carried out by Nature involving 1576 researchers states that ‘more than 70% of researchers [involved in the survey] have tried and failed to reproduce another scientist’s experiments, and more than half have failed to reproduce their own experiments.’
But does this mean that most of the published research is wrong or untrustworthy? Not necessarily. The same study suggests two reasons behind these reproducibility issues:
- the complexity of science: sometimes slight modification in the conditions in which the experiment is carried out might make a huge difference
- Selective reporting: cherry picking data in support of a hypothesis, for instance.
When asked to select from a list of potential solutions, 90% of the Nature survey respondents choose better mentorship, proper use of statistics, and more robust experimental design.
Synthace Ltd., a UK based technology company is already working on the development of a platform that could help ‘biology engineers’ solve the problem by providing the means to interact with lab equipment and automation in a much more flexible and high level way than is currently possible. ‘Automation is key because of both its intrinsic repeatability and the experimental bandwidth it provides to do better designed experiments,’ says Tim (CEO of Synthace).
When Synthace was founded in 2011, it began life as a synthetic biology company looking to exploit the huge power of biology to manufacture novel products. However, they soon realized that they did not have the necessary tools to do this. They, therefore, proceeded to invent those tools, and in the process they came to understand that the platform they developed could be invaluable for the entire community. Synthace is now a software company for biology, focused on the development of a transformative technology that will allow whoever wants to engineer biology to do so in a much more reliable way and exploiting the power of biological complexity more thoroughly (for instance, considering many more experimental variables at the same time).
Synthace’s technology platform is Antha.
‘Antha is a high-level programming language for biology and an operating system for all your laboratory hardware, making it easy to design and crucially, actually execute, experiments which are reproducible, reliable, and fast. Antha allows you to abstract away many of the complexities of working with biology by linking software and hardware with laboratory working practices.’
Tim Fell and Sean Ward (respectively CEO and co-Founder/CTO of Synthace) will help us understand their system, work and objectives in a Q&A session.
Daniela: can you tell us more about the idea behind Antha?
- Sean: Antha comes from the realization that, in a sense, everything we do in the lab can be software driven. Until now we have been using people [to set up and run experiments], and we want to transform this into an automated process, freeing up those highly trained scientists to use their time and insights more valuably.
- Tim: Biology is extremely complex: even the simplest cell is far more complex than the most sophisticated artifact that we have ever made and yet we are trying to address that problem space with a hand-held pipette, pen and paper and an Excel spreadsheet. We should not be handling this kind of complexity manually: we believe that we need a completely different set of tools to work with this complexity, to deal with repetitive operations and to be able to do things at the right scale. When you are trying to develop a technical solution you’ve got to have enough experimental bandwidth to address it properly. This is why we developed Antha, in order to provide a flexible interface to lab automation and analytics. We want to make the technology available to people that are not expert in robotics, without the need to code to make the instrument work. You need to be able to just say ‘I want to assemble these thirty [biological] constructs’ and engage with the language that biologists understand. When you are working with Powerpoint in Windows, for instance, you don’t care how it works behind the scenes. That’s what we have written, that kind of operating system that allows people to easily specify what to do [without the need of possessing hardcore programming skills]. You don’t care if you have a Canon or an Epson printer: what you care about is that, once you press print, it is going to work. Since much of molecular biology involves moving small volumes of liquid around, liquid handling robots [that are able to pipette liquids in different amounts and with extreme precision] provide huge benefit, so we are writing Antha drivers for the producers of liquid handlers (e.g. CyBio® or Gilson that are making the smallest and most affordable liquid handlers, or Tecan and Hamilton). This will give people access to a whole host of what we call biological unit operations, which they can then simply assemble using Antha’s graphical interface into experimental workflows to execute on their equipment regardless of the brand.
Daniela: How does Antha allow you to ‘unlock biological complexity’?
- Tim: you have to look at biology as the complex multicomponent system that it is. You can’t be reductionist to the extent you ignore the higher order emergent interactions. Antha is a tool by which you can now experimentally access much higher factor spaces and unpick those emergent effects. It enables an iterative approach in which we can select many factors and do a holistic observation of them, build a model, and iterate through another round of empirical experimentation.
- Sean: We’ve expanded our use of machine-learning heavily to build models that help us understand very complex [multifactorial] systems. DOE (Design Of Experiment) was our initial framework. Generically we use a branch of machine-learning called active learning. Antha as a language was designed to implicitly enable the improvement of any sort of experiment or working practice within it, guiding you through deciding what is the factor space that makes up a good working practice the same way you think about the factor space of something like an entire biomanufacturing process. The AnthaOS then makes it very easy to do these high-dimensional investigations, to optimize and characterize the reproducibility of your working practices, rather than only thinking in terms of the biological problem. Because, of course, if the base working practices which you are conducting your experiment with are themselves unreliable, they are contributing to noise and error that propagates to the actual experiment, masking what is actually going on in the system.
Daniela: What is your final goal and at which stage are you today?
- Tim: We envisage a world where all researchers will be able to download other people’s highly optimized and validated working practices at the click of a button, and not have to go back to that Nature paper to try and rework that protocol with the lack of information that is intrinsic of most papers (every scientist knows what it is like to have to spend many weeks trying to optimize somebody else’s protocol!). Then to be able to rapidly assemble any new experiment from these validated elements and automatically execute them on their own equipment, regardless of make or manufacturer.
- Sean: A large number of our standard library elements are already published and available today (people are already using the open source version of Antha). An important thing to understand is that what Tim is discussing is our longer term goal. Today, you can describe a working practice in Antha in 5 minutes. It is the same as writing an SOP in English once you are proficient in the language. However, for experimental validation that your SOP is robust and reproducible it does require actually performing/running the experiment and confirming that it is in fact a reliable working practice. This is an important distinction to make. It does not take a massive amount of energy to use Antha in general R&D experimentation, in fact it is already faster than how we work today. However, it does take energy, with real personal and communal benefit as a result, to build a reproducible working practice.
- Tim: The Synthace team is now building libraries of Antha elements and validating those operations in the wet lab in the way Sean describes. This is absolutely central to the way we work.
Daniela: Who are your clients at the moment?
- Tim: The announced clients that we can talk about are Merck (known as MSD in Europe), and Dow AgroSciences, and we are at contract stages with a several other multinationals.
Daniela: What about the academic laboratories? How likely is it that they will start to use this new technology?
- Tim: Very likely. This is the reason why we first wrote Antha drivers to CyBio® and Gilson instruments: these machines cost between fifteen and twenty thousand dollars, which is well within the reach of a Principal Investigator in academia. When you realize that you can make a thousand genetic constructs in a week on one of those machines with Antha and that they all work, this suddenly gives you a way of accessing biological experimentation that was just unthinkable before. Academia is definitely a target market for us. However, targeting the large companies has been our priority to start with. They are well resourced and they have a huge base of installed equipment which they are not utilizing at their full potential. In fact, the utilization of biotechnology equipment is depressingly low: I dare say under 10% in many laboratories. It is common to hear of pieces of automation that were bought for one particular experiment which took somebody several weeks to program it for (it might have been a high throughtput screen, for example), but once that experiment is over, these things have a habit of standing idle. What you really need is some way to be able to reprogram them very flexibly and quickly to do different experiments. And this is what Antha allows you to do. You can conceive a new experiment, design it and implement it in a matter of minutes. In a lot of big companies, people are thinking this way: how can we get more utilization of these machines? How can we get more reproducibility? How can we make our experimental design more robust? Do bigger experiments? They see that they have the equipment but they can’t interface with it. They can’t engage with it. They are very obvious first customers of ours.
Daniela: If you were an academic lab and you were interested in using Antha, what would you need to do? What do you require?
- Sean: It works most effectively if you have access to a liquid handler, of course, and the most progressed integrations are with the CyBio® or Gilson hardware, at present. We will announce a number of liquid handling systems supported by Antha from other manufacturers this year. It is also possible to make Antha generate human instructions to tell you manually how to perform the experiment. If you are talking about a 5 parts assembly for 32 constructs that’s about 1700 liquid handling instructions: you don’t want to do this by hand. The future of biotech is not having a lot of people getting repetitive strain injuries.
- Tim: To take advantage of the full utility of Antha, you want a visual user interface (UI) that makes it easy to use, and you need to purchase a license for that part of the system. Antha has been a technology in our own lab for many years but what we’ve achieved recently is to productize it, so that it is now very easy for people to use from a visual point of view.
Daniela: Can you share with us your vision of the synthetic biology field (and of biology engineering in general) for the future?
- Tim: Biology will be our major form of manufacturing in the future. We will heal, feed, manufacture with biology. [Nature] is so sophisticated, elegant and powerful. It has to be the future. We have done incredible things with it for thousands of years (e.g. vaccines or agriculture), but we can do so, so much more.
- Sean: There is one major thing happening in the world at the moment. As working with the physical world is becoming increasingly digital, every company that is out there is discovering that they either are a technology business or they are dead. And that is what is happening with biology: it is becoming a technology business.
You can follow Synthace on twitter @
Disclaimer: Daniela has worked for Synthace in the past (2013/2014), but she declares no conflicting interests at present.