The Center for Data Innovation spoke to Kristina Lagerstedt, co-founder and chief executive officer of 1928 Diagnostics, a Swedish startup that provides an online platform researchers can use to analyze genetic information in order to predict antibiotic resistance and identify effective antibiotic treatments. Lagerstedt discussed how data analysis can help antibiotic-resistant bacteria and prevent outbreaks in hospitals.
Nick Wallace: Antibiotic resistance is a growing problem. What is the root of the problem, and how can whole genome sequencing help?
Kristina Lagerstedt: We can start with our name. The scientist Alexander Fleming discovered antibiotics in 1928. Since then, the average lifespan increased by ten years. But in 1945 he received the Nobel Prize, and already at that time, he mentioned the risk of bacteria becoming resistant to antibiotics due to misuse of antibiotics: that is, not treating the patient for long enough or with the correct dose.
So we started 1928 Diagnostics to save the power of antibiotics. Our contribution is a software as a service (SaaS) for diagnostics and infection control, by using next-generation sequencing, which is a technology that reads the whole genetic chain of bacteria or any organism. What we do is enable the hospitals to use this sequencing technology for diagnostic purposes, for monitoring bacteria, and to prevent outbreaks. So it’s the perfect genetic fingerprint, but we need to uncover all the information that’s biologically and medically relevant within the genetic code, and do that within like three minutes, by data processing and comparison to our databases.
There are two sides to how we address antibiotic resistance. One is decision support. That’s where you reveal the genes and mutations in the bacteria that cause antibiotic resistance, and if you have knowledge of this, you understand what antibiotics you cannot use. On the other hand, we have the infection control approach, which means that you use the whole genetic chain of the bacteria to identify it, so you can tell if you have seen it before and how closely related that bacteria is to another found in your hospital a week ago. So it helps you to prevent outbreaks and to control outbreaks, and to control bacteria and erase the bacteria that you have in a hospital.
Bacteria are smart, they’ve been around for billions of years. How did that happen? It’s because they found a way to survive. By surviving antibiotics, which are supposed to kill them, they create systems and proteins that break down the antibiotics and flush them out and they need genes to do that. They can even share genes with each other: an E-coli bacterium can share genes with a Klebsiella bacterium. They’re really smart, so we need to be as smart as they are in uncoding their genetic code.
Wallace: What data does 1928 Diagnostics use, and how does it provide clients with the answers they’re looking for, such as the probability of antibiotic resistance?
Lagerstedt: When we started this company, I was working as a cancer scientist, working with lots of molecular biology methods and generating huge datasets. I couldn’t do the analysis myself, it was too tricky, so I started collaborating with Erik Kristiansson at Chalmers University of Technology. Then I worked with children with pediatric acute lymphocytic leukemia (ALL). Ten percent of these children die, so the survival rates are high, but among those who die, they don’t die from the cancer itself. They die because they get bacterial infections that cannot be treated fast enough with the correct antibiotics.
Then we realized, that Erik had a database of genes that cause antibiotic resistance, and a bunch of algorithms. So we thought that if we could put this together to make manually-curated, high-quality databases, based on validated evidence from published results from research studies and peer-reviewed articles, and then to automate the algorithms, put this together in cloud-based software and make it fast and scalable, then we could provide an ability for the users to just upload raw data files from raw data sequencers. There are mainly just two brands of sequencers that cover 85 percent of the market, and they have the same format, called FastQ. These FastQ files contain data on nucleic acids, or just the four components that make up the whole genetic code.
We get the raw data, not processed in any way, and then we process it with quality algorithms that compare it to our databases on antibiotic resistance and something called cgMLST, which is to identify what kind of bacteria it is, and then within less than three minutes they will get the results back.
Wallace: Do you have any plans to use artificial intelligence? What impact do you think that technology might have on this type of research?
Lagerstedt: This is the interesting thing: we cannot say we’re using artificial intelligence right now exactly, on a regular basis, because what is artificial intelligence? What is machine learning? Maybe we could say we’re doing that, but we are very serious about the meaning of those terms. We are setting up projects, and we will soon start a collaboration with a New York based bioinformatics group in the machine learning.
The first things that comes up is, “can you predict antibiotic resistance by just reading the genetic code?” When we asked people a couple of years ago, they said no. Now research groups are already doing this. But you need to do it in a validated way, so that’s what we’re working on, because that’s the difference between being a scientist and being a company, because if the hospitals are supposed to use this, it must be validated and secure and quality-assessed. So that’s the first thing: to predict antibiotic resistance, not only by comparing it to data that is already published, but also to predict it on genetic changes that cause resistance that have not yet been identified in a genetic study.
The second thing is to predict outbreaks. Predicting resistance is in the domain of diagnostics, and predicting outbreaks is in the scope of infection control. Those are the two key directions we’re working on right now.
Wallace: Antibiotic resistance ranks pretty high on the “list of things to worry about.” Since you’re much closer to the science, how do you see the prospects for solving the problem?
Lagerstedt: Antibiotic resistance is in the top three threats against humanity according to the World Health Organization, parallel to lack of clean water and the climate threat. I think that collaboration is the key here. That is to share data, and that’s what we’re working on: to be able to share data between authorities across borders, between hospitals, with whoever you want to share your data. Not only the genetic data, but also the metadata that is connected to it. So that is what we’re working on right now in pilot projects, and a scale pilot project that we started last year with hospitals in Sweden. Now we are expanding it to working with authorities in Europe, and we are also going to scale it into the United States.
That’s why it’s such a privilege to be awarded the tech pioneer award at the World Economic Forum, because I can talk to decision makers all around the globe, and really to leverage and point out the problem of antibiotic resistance on a global scale and to be able to be part of this together with all these fantastic people, it’s such a huge privilege so we really take this as an opportunity to work really hard and solve this problem through collaboration
Wallace: Are there any regulatory or policy obstacles to the kind of work you think you could do?Lagerstedt: Yes, there are a lot of considerations. This space is totally new. Going to the Food and Drug Administration (FDA) and saying, “hey, we’ve got databases here and we want to do AI,” that will take a while, and we’re a small company. We have a CE mark (a health and safety certification of the European Economic Area) on our first product for decision-support for MRSA, and we have custom-made quality control systems. But we still need to work closely with authorities like the FDA because no one has done this before.