It is no secret that artificial intelligence is currently altering many sectors more than ever. According to experts, AI in healthcare is fundamentally transforming patient health. For the life science company Bayer, it is particularly helping doctors to better and earlier detect diseases, to custom tailor treatments to individual needs, and also to solve complicated health issues by developing and producing medicine faster. All this is possible because of the available data which can be generated and interpreted by data scientists through AI. Dr. Angeli Moeller is a professional data scientist who speaks data fluently and has been working with Bayer for almost three years. She is leading a team responsible for digital investment in pharma research as well as an artificial intelligence program. She formerly worked as a molecular oncologist on cancer research in the UK, as a system biologist in Berlin at the Max Delbrück center, and as a data scientist at Thomson Reuters.
When was the first time you bumped into AI?
I am, and always have been, very much attracted to things that are cutting edge, things that are new but also very difficult and challenging. So I started off in genetic and stem cell research, which was very new when I was 18, and later I developed nanobodies during my PhD. All of this research required I analyze data and my skills grew with my datasets.
As a postdoctoral researcher, I was on a European-funded grant – or actually two grants – one called SynSyS and one called Eurospin. Our research focused on neurodegeneration and memory formation at the synapse. We worked in collaboration with 15 different labs, one of which was headed by the Nobel Prize winner Erwin Neher. The data was large and varied and only through the power of machine learning (a.k.a. artificial intelligence) were we able to make meaningful biological predictions which we subsequently validated in the lab.
“I have always wanted to work in health care. We have a lot of important open questions on diseases that we still have to tackle.”
What triggered you to go into this sector?
I have always wanted to work in health care. I think in health care we have really important problems to solve, a lot of important open questions on diseases that we still have to tackle. And I want to make use of the resources that I have, whether that is data or a new type of mathematic solution to solve those problems. And machine learning has brought me closer to that.
You are currently working at Bayer leading a team responsible for digital investment in pharma research and also the artificial intelligence program. What does your job entail?
At Bayer, we have what we call an IT business partnering organization. We are accountable for the budget, which we spend on changes in IT and digital infrastructure. To give you an example, we have a project right now where we’ve invested in the use of artificial intelligence for chemical synthesis. And this is trying to model the best structure of a new drug we want to develop.
Can you break all that down for me to easily understand?
We are supporting the discovery of new medicine with IT and data-science solutions.
“We have three main areas in which we operate: oncology, cardiovascular research, and women’s health.”
For which disease?
We have three main areas in which we operate: cancer/oncology, cardiovascular research (everything about the heart), and women’s health (e.g. contraception). However, during research and development we really focus on how the drug works at a molecular level, which means we can identify drugs that have an impact on more than one disease. One key factor in deciding which diseases to target is, “Where do we see the most need?” For example, Bayer developed Jivi, a preventative treatment for the rare bleeding disorder hemophilia A. The decision to develop this treatment was based on the identification of a disease area in which patients previously had only had limited access to medication that could make a significant improvement to their lives. We call that a high unmet medical need.
I have seen people die of cancer, not because they are not treated, but because they are taking drugs that don’t react to their bodies or cells as planned. Is this also your area of research?
Absolutely. Touching on that personal experience, it is something which is very common for us here. Cancer and heart disease are part of most people’s lives, affecting friends and loved ones, and that includes the people who are working on these projects. Almost all of them have some kind of personal experience and it’s often what motivates them to work in the healthcare sector and not in a different industry developing cars or something else.
But specific to your question, there is one project that we’re running in collaboration with Turbine, a company that came through our G4A program for collaboration with startups. They are working on models of how cancer works on the molecular level and tests millions of potential drugs on it with artificial intelligence. And it is exactly because of what you pointed out; in oncology, in cardiovascular disease, and in many other areas, people have a whole array of different medications and environmental factors which impact how drugs behave. Artificial intelligence allows doctors to model that complexity by recording data on what you eat, when you go to sleep, and what medications you are taking. This helps doctors to tailor medications to each individual patient instead of treating a person as an average person in a population that has that disease.
“Complex predictions regarding human health are enabled by machine learning – and are creating new opportunities to develop better treatments.”
Has AI made your work easier when it comes to dealing with these issues?
Yes, complex predictions regarding human health are enabled by machine learning – and are creating new opportunities to develop better treatments. If you are a training molecular oncologist or a doctor, you start by memorizing which things affect which disease. But you are sometimes limited by how much information a human being can process and make sense of; this is where you can turn to machine learning/artificial intelligence. Artificial neuronal networks, which are part of the artificial intelligence tool-kit, allow us to make meaningful predictions even when we don’t know all the variables. Artificial intelligence is trying to make sense of the masses of data now collected in daily life and at hospitals, enabling more precision and accuracy in medicine.
I gather that AI is really important in the health sector. What do you say to people who say that AI is a problem or a danger to human beings?
With any new technology, machine learning included, you can’t and shouldn’t separate the ethical discussion. We need a broad and informed debate on the opportunities and risks of AI in society, and Bayer is heavily engaged in this debate. For example, we are part of the High Level Expert Group on AI of the EU Commission, which is working on an ethical framework, and I personally sit on the executive committee of the non-profit Alliance for Artificial Intelligence in Healthcare (AAIH), which is also working on ethical questions of AI.
You touched a very important topic there, ethics in AI. Many people are criticizing that AI cannot be ethical because AI is a machine and cannot be trustworthy. How can you prove to me that anything you do to me by using AI is trustworthy?
Machine learning, or AI, is a technology that can support a decision still made by a doctor. AI is just a tool that provides additional information, which in the past was on a dashboard with some very simple Excel-based analytics in the background. The machine allows the doctor to make a more informed decision. But in the end, it is still that healthcare professional’s decision that they make together with the patient about the best treatment option.
How can data science impact the health industry?
Consider all the people who are now entering health data on their smart watches, on their mobile phones, plus all the electronic medical records that the hospitals are collecting during a patient’s life in different countries. There is just a lot more data available and I think if it is correctly leveraged, that data can give us the opportunity to have a much more accelerated impact in areas of high unmet medical need. Artificial intelligence makes it possible to move faster in targeting these incredibly challenging areas of medicine where we as a society, for different reasons, haven’t made the accelerated progress we would like to see. It is a new opportunity to tackle these most difficult healthcare problems.
“I have always wanted to be a scientist. I have just gotten my seven year old a new science kit for her birthday and I hope she wants to be one, too.”
Have you always dreamt of being a data scientist?
I have always wanted to be a scientist. I have just gotten my seven year old a new science kit for her birthday and I hope she wants to be one, too. Healthcare was definitely the sector I wanted to go into for a long time. I think the good thing about being a data scientist in healthcare is that, maybe not in a dramatic way, but in a small way you can feel like you are doing the right thing. That you are making a contribution. And that is something that I enjoy, even though I am perhaps not changing the world, but every day I feel like it’s a great contribution. It is something you can go home and feel good about.
What is it that you are really proud of in terms of projects?
Right now, what is really on my mind is the oncology project that we are working on. We have a drug for lung cancer that is only effective on patients who have a particular genetic mutation. Only one in 300 patients with lung cancer has the mutation and would therefore benefit from this treatment. Together with Auria Biobank in Finland we are developing an artificial-intelligence-based algorithm that can identify those. I especially like this project because of the speed – we started off with the concept over Christmas and have been able to move very quickly and develop a minimum viable product within five months. I find that very important in a big company like Bayer to see that we can be agile and move quickly towards having impact on patients.
What is the most important lesson you have learned along the way when it comes to working with AI?
The most important lesson for me is to listen to the experts and to have a mix of expertise that you need in a project team. I am proud to work with such talented teams at Bayer.