Tobias it is a pleasure to meet you, could you give our readers an introduction to yourself?
Certainly, my name is Tobias I am originally from Bremen, and I founded Aigenpulse around 2 years ago. My background is in Life Science research. I did a PhD in Bioinformatics back in Germany, before moving to do my postdoc in Cambridge and Oxford. I initially worked at the Laboratory of Molecular Biology in the research group of Sean Munro and later with Matthew Freeman at the Dunn School in Oxford until about 4 years ago when I moved out of academia and into the private sector.
Around that time data science rose to prominence, so I did some initial consulting work for a few years and ended up building a consulting company with a business partner; now Aigenpulse is the sole focus of my work.
What motivated you to pursue a path in Life Sciences?
My background is actually theoretical Mathematics and whilst doing my undergrad I started working as a system administrator at the Max Planck Institute for Biophysical Chemistry in Germany. I was fascinated by how much attention to detail was required and how much time scientists would spend trying to create insights from complex experiments. I think it is mindboggling how much you can do with biology these days, so that really piqued my interest. I subsequently moved further away from theoretical work and more into collaborating with biologists, helping to combine their biological knowledge and intuition with insights from data.
Could you give us an insight into Aigenpulse?
I think with the advancements of data technology over the last 10 to 15 years, it became clear that if you used data correctly, you could actually help in a lot of fields within biology. Sequencing is a very good example of where we have seen a rapid improvement in the technology. What was achieved with much effort 10 years in ago in this space, we can now do routinely.
The challenge with that is that there is a lot of information created and therefore proper systems have become essential to help manage that and derive knowledge from it. Aigenpulse was created to help with that; we see ourselves as a company that uses data technologies and machine learning to improve research outcomes.
We have a software platform that helps to structure data, which then enables the use of machine learning to automate analyses and to see trends and associations in data that wouldn’t be apparent to the human eye.
You mention how quickly technology has advanced in the past decade, specifically within Healthcare, could you try and predict one of the big breakthroughs you expect over the next few years within Healthcare and AI?
That’s a very good question. Interestingly I don’t think it is just the technology advancements that will bring the biggest breakthroughs, it will be when we bring technology and biology together.
For example, we can see a very big drive towards personalized medicine; however, there are difficulties in being able to efficiently develop and administrate personalised drugs. Being able to correctly predict treatment outcomes quickly and accurately will need advanced machine learning and AI. That’s where we can have a huge positive impact for the population.
Is there a part of healthcare that you don’t you think AI can impact in a positive way? is there a part of healthcare that will always need a human touch?
In the next couple of decades, I think there will always be a human touch in healthcare. I cannot, currently, foresee a future where there are fully automated systems running a nursing home for example.
The benefit of trying to achieve this is, for instance, if you look at the cost of full-time care for Alzheimer patients in the US alone, the number is massive; way into the billions. I think being able to streamline the cost, and also improve patient care with AI can really help with that, but I cannot see AI fully replacing humans in a full-time care capacity any time soon.
We saw recently that an advancement AI was able to prevent Breast Cancer 99% of the time. This was great because I think we should be seeing positive news reach our society, as I don’t believe that the general public are aware of just how beneficial AI can be in our Healthcare system. Do you think that we are doing enough to educate people about what their data is, or can be used for?
I am an advocate for telling people, for sure. Now the question is what is ‘your’ data? If you go to the hospital and they do a test on you, is that data yours? The hospitals? The Trusts? The insurance company who paid for the test?
Ownership of data is still something that is not completely clear in healthcare, and that alone would be an interesting discussion. But if you look at how data is being used, a lot of the information that is being used is donated data, which is hosted in public spaces, for example the 100,000 Genome Project from Genomics England.
Whose data do you think it is?
I think it is the patient’s data. How we would ensure that we find a balanced approach and what technology should be used to ensure that that happens moving forward is an interesting topic. There are some interesting technologies that will be coming to the market which may be very helpful with that, but it is an unsolved challenge. So, whilst we acknowledge data should belong to individuals, there currently is no technology that I am aware of that guarantees that.
From the discussions I have had previously, Healthcare has been highlighted as an industry that really pioneered the use of data, why do you think that is?
I think it is because a lot of people who work in this space come from a background that uses data, for example, in research. So, they are used to data, and how useful it is; analysing the data in tools like Excel, they understand the basics of data.
What becomes more interesting nowadays is that you have multiple large data assets which become more and more useful when they are interrogated together. And that is actually one of the core challenges that we are addressing with our platform.
I believe, because everyone in healthcare generally has a great appreciation and understanding of the usefulness of data, efforts to integrate disparate data sources is an accepted challenge. Discussions today are more about how you get there, how you make integrated data useful, what are the next steps? We spend a lot of our time discussing with peers on what can be done with machine learning and AI.
The other challenge is that, in research, technology is evolving continuously. There is a new technology, that can have a big impact, being developed every year. If there is new technology that helps create more data or use data differently, you need to be able to adapt to that. It is a big challenge to build systems that can continuously integrate with new technologies.
What are some of the key challenges faced working with Healthcare Data?
In my mind, most technical transformation projects have two core challenges:
- The technology challenge
- The political challenge
Because change is always political!
These 2 challenges are always there in life sciences, but there is one extra challenge in life sciences and that is the rapid change and development of the research. This additional complexity makes technical transformation projects in the life sciences especially challenging.
How has 2018 been for Aigenpulse?
One of the biggest highlights is that we have grown quite a bit. We reached a very interesting point late last year that enabled us to grow the company 4-fold. That has been very interesting. I have been very fortunate in that we have hired some outstanding people, it is one of the areas I really focused on. Knowing how to hire great people is the backbone of our growth.
Other things I have enjoyed is working closely with our customers, as they have been so supportive of our journey; we haven’t actually raised funding, as such. It has been a very cash-flow driven company which has its own pros and cons. But that would not have been possible without our customers!
We have just opened an office in the US as well. There is a big life sciences cluster in Boston, and we have been travelling there regularly to build our network.
Technology wise we are getting to the point where our product has matured, which is exciting. We have integrated the majority of the complex data sets our customers have, so we can shift our focus to building the structure of the data that we need, in order to optimise the use of machine learning. Until now we have spent 80% of our time ensuring that the data structure is correct and 20% delivering machine learning. I think that will shift in 2019 quite dramatically and our focus will become how you provide AI and machine learning on these very large complex data assets; that is very interesting!
We have access to public data assets and we combine them with the client’s data assets, so they can be mined for interesting targets, better understanding patient risks and effectiveness.
What is the best thing about working within Healthcare & data?
For me this space is ripe for better usage of information and data.
I think the sector has been held back by the third challenge we discussed previously, the challenge of the always evolving research technologies. What has been very exciting to see, with the rise of the continuous integration and continuous deployment pipelines, is the ability to evolve a platform alongside that kind of research projects That seems like a really interesting opportunity for me.
Nowadays, you can build a product that evolves alongside research while research itself advances. I wanted to build that, because it seemed to be missing. If we can further mature our product then we will have a huge impact on the success of the drugs being developed. It should also ensure that drugs are cheaper and more accessible, because, if we can make outcomes more predictable, then the money you need to spend on a new drug will be reduced.
What tech are you guys utilising at the moment?
We mostly use Python Django, for the user interface. For the data management we use anything that we think performs well enough. For the machine learning environments, we use frameworks like H20 & TensorFlow. We are very open to using the best technology to solve a challenge, rather than tying ourselves down to one specific solution. Most of the product is a structured database; going forward we are exploring graph databases for example. A lot of the larger genomics data assets that are being processed for variant analysis cannot be stored efficiently in a relational data structure, they will have specialised databases.
So, bringing all that together in a manageable, scalable and compliant environment is technically quite challenging and interesting.
Is there anyone in the AI Healthcare space that you think is pioneering the field?
In my mind, it is BenevolentAI. I think they are really becoming true to the vision they set over the last few years. What they are doing is exceedingly interesting. It is a different take on how life sciences could work, by putting a technology at the core of every decision. It will be hard to get to the vision behind it, but they are doing a great job.