How can we work more sustainably, smarter and healthier? In this edition, expert lecturer Data Science for Life Sciences and Health at Hanze University of Applied Sciences Groningen, Wynand Alkema, talks about the importance of data research for humanity.
Two days a week Wynand Alkema's life revolves around data research. At least: data research for science. Because the Data Science for Life Sciences and Health lecturer is busy the rest of the work week with TenWise, the company he founded in 2014 that essentially does the same thing: making chocolate out of (bio-)data, but for the benefit of business.
'You are amazed at how much data has already been collected in numerous studies worldwide,' says Alkema. 'The trick is to discover connections in it that lead to new discoveries. And that is what we do. We use algorithms to see patterns in all that data. And that benefits physicians, researchers, pharmacists and also food companies. We provide new knowledge by combining existing knowledge. This is how you should see it. This is AI. It's fashionable now, but we've been using it for years.'
As an example: for many years, many studies have been done around the world on colon cancer. Those studies have tracked how certain proteins behave in specific patients. Alkema and his students put those results side by side and discovered a pattern. Say: protein 538 and protein 10372 are activated and then things go wrong in the body. That's extremely valuable information for medics designing a drug, of course.
Centimeters and inches
'We use mountains of data that are freely available for such an analysis. They come from public databases. Look, after every scientific study, the data have to be released. We know where to find them and can put them into our algorithm. That applies to a lot of data, almost always anonymized. About blood pressure, substances in urine, you name it. If you put them together smartly, you can discover important patterns. These say something, for example, about why some patients live longer than others.'
The number of datasets available online is staggering. So are the opportunities to gain new knowledge from them. The trick is to get the data "clean. It has to be reliable and allow the data scientist to make good comparisons. This takes up a lot of time for the 150 or so students who are involved in this at Hanze University. Alkema: 'Those studies are done all over the world, in all languages. You don't want to mix up centimeters and inches, so to speak. That has to be equalized first.'
Those so-called numerical data are one type of data that students use. Another is literature. All those researchers leave behind huge amounts of reports, records, dissertations, articles and whatnot. There is valuable information to be found in each, which can be combined - again - to arrive at new insights.
Wynand Alkema explains: 'One researcher writes about how vitamin C works well to stimulate protein X.' Three years later another researcher sees something similar in a laboratory animal, and two years after that another discovers the same thing in a bacterium. They don't know that about each other, but we discover that connection. We have access to 35 million scientific articles. You can bet that patterns and connections can be found in those.'
Millions of cookbooks and recipes
This type of data does require an extra step: reading those millions of articles. A clever program does that these days. 'That program reads PDFs and makes all those words searchable. That's what we need. Then we can ask the computer to look for - in this example - combinations of the terms vitamin C and protein X. When there is enough data, we can draw conclusions. We are also building a knowledge network in this way that researchers and developers in Life Sciences & Health can use.'
We use algorithms to see patterns in all that data. And that benefits physicians, researchers, pharmacists and also food companies.
Wynand Alkema
To keep the data pure, Alkema and his team train the algorithm continuously. 'We give a list of facts that we are sure are correct. We also tell it which thousand are absolutely wrong. In this way, the algorithm learns what to pay attention to and the results constantly improve.'
By conducting the process properly, valuable information can be obtained. The medical knowledge being expanded is substantial; the opportunities for useful follow-up research are great. 'If we have enough useful data, we can predict the effects of certain drugs on specific patient groups. That's as far as it goes.' But the Hanze University is doing more. The food industry also benefits from the research.
Alkema: "We investigate which substances can best be used to make plant-based meat substitutes, for example. Then we look at cookbooks, online recipe sites and especially all the reactions of users. And so, by combining millions of data, we can predict which spices provide the right smell, which substances best approach texture, and so on. We can do the same, of course, when it comes to the nutrients needed for people's health.'
Order
The step now is: what do you want to know? We extract it from the data. That step sounds obvious, but it's not so easy in practice. "Of every project, we spend about forty percent of the time consulting with the questioner. We can only get to work properly if we know very precisely what is being asked. That has to be precisely delineated so that we can deploy the algorithm effectively. And that's quite tricky. Do we want - to stay with an earlier example - vitamin C combined with protein X? And should we then omit outcomes with protein Y or not? And with that meat substitute: do you need the smell of meat that most people worldwide liked, or are you more interested in the smell that is nice for Europeans? Those are just small examples.'
Talking is it, asking questions and finding out exactly what a researcher, an entrepreneur, a developer wants with the data. Alkema: "We train our students in that. That's why it's also good that they have a background in Life Science & Health. It is essential that they understand roughly what the questioner is talking about, so they can frame it better.'
This type of research based on artificial intelligence is on the rise. Logical, because the data are there and much can be learned from them. 'We still have a world to win in that respect. Organizing what is already there is what we are doing.'