The Lab meets smart algorithms
Mark Hanemaaijer has been Winclove Probiotics’ bioinformatician since 2018. His work involves ensuring the safety and effectiveness of the strains that go into our formulations. To feed Winclove’s unique, ever-growing database, Mark gathers scientific knowledge by means of DNA analysis and big data. In this interview Mark elaborates on this crucial fusion of lab research and the latest technological developments.
NAME: Dr. Mark Hanemaaijer
POSITION: Bioinformatician, Strain Development Department, (subdivision of Winclove Probiotics’ Research & Development)
EDUCATION: BSc Biotechnology, Delft University of Technology / Leiden University; MSc Biotechnology, Delft University of Technology; PhD Systems Biology, Vrije Universiteit Amsterdam; post-doctoral research in Population Genomics, University of California at Davis.]
About Mark’s 2018 move from the world of research to that of business, he says: ‘I really wanted to go into the business sector. There are about 100 people working at Winclove, and the great thing about that is that you get to know what is happening on the research and development side of the company, but also on the operations and marketing side. To me that’s really an added value. When you do academic research, your focus is so narrow. But I like working on a variety of projects and subjects, and Winclove gives me plenty of space for that.’
Safety & Quality
Mark’s work is innovative: he explores potentially interesting strains for new formulations. He says, ‘I work a lot together with my colleagues in the lab, and basically I research the properties of not just our own probiotic strains but of other bacteria that might be interesting too. To do that I look at the DNA and sift through large data sets. The information that we obtain from that contributes to the safety and quality of our formulations, but it also helps us discover new directions for strains or functionalities.’
A unique strain database
The Winclove R&D team maintains a safety dossier for each of its strains . Mark explains, ‘The team members check for example whether a strain doesn’t make any toxic substances and isn’t resistant to antibiotics. We guarantee the safety of the strains we use in production. The team also inventories properties that we think could be useful for our formulations. For example, we determine whether a strain survives in the gastro-intestinal tract, what substances it produces, and what the influence on the intestines and immune system is. In this way we’ve set up a unique knowledge database for our strains that we continue to update. So afterwards we can refer to all this info when we want to improve an existing formulation or develop a new one.’
Supporting the product dossiers
And as if gathering the knowledge for the development and improvement of our established and developmental formulations weren’t stimulating and productive enough, Mark also handles technical enquiries from our business partners. He enjoys working on questions such as: ‘Could our bacteria do this?’, ‘How similar are our bacteria to another bacterium?’, and ‘Substance X shows a certain mechanism in the literature; can our bacteria make that substance too?’. With the information that Mark amasses in tackling these questions, we then compile even stronger dossiers for our partners’ formulations.
Mark’s expertise: DNA analysis
Prior to the advent of DNA sequencing, lab testing was the sole generator of knowledge of strain properties. Mark recalls, ‘You had to cultivate each strain separately and do biochemical tests to determine the functional properties of the strains. That was the only way to discover that.’ Not only was the process very costly and time-consuming, but some bacteria were difficult to cultivate and therefore excluded from probiotic innovation. While this remains the case even today, by using the bacterial strains’ DNA we can now assess their properties without the time demanding task of cultivation. And there lies Mark’s expertise.
Roughly fifteen years ago it became easier and more affordable to sequence the complete genome of living creatures. Consequently, information on the DNA of many bacterial strains is now accessible in public scientific databases. Mark says, ‘Winclove too has mapped our strains’ DNA. This sequencing opened the door to the linking of properties to the DNA, or, to be more precise, to the genes. When you look at the properties of the strains, then you see that certain properties co-occur with certain genes. And then you know it’s highly probable that a certain gene is responsible for a certain property.’
That linking of DNA to a property happens not in the lab, but with sophisticated computer algorithms. These algorithms are indispensable for comparing DNA quickly. Mark explains, ‘The DNA of our bacterial strains contains a hefty 1.5 to 3.5 million nucleotide pairs – that is, the combinations of the substances adenine (A) with thymosin (T) and of guanine (G) with cytosine (C). It would take months to compare them by hand. Whereas with the computer power of nowadays, these sophisticated algorithms do it in just minutes or hours.’
On top of this, Mark applies Artificial Intelligence (AI) algorithms to innovation. The highly advanced algorithms involved can independently reveal connections between data sets. Mark says, ‘The outcome is a mathematical model that estimates or predicts what strains, properties, or genes are linked to certain outcomes. Eventually, I want to be able to use the data to predict what strains, properties, or genes have a favourable effect on the body.’ Examples of the feasibility of this exist: With such algorithms we predicted whether certain strains in our strain database were resistant towards antibiotics by predicting the Minimum Inhibitory Concentrations (MIC) for 10 different antibiotics. Using it, Mark has analysed Winclove’s strains to a degree of precision that until now was well beyond reach.
Bacteria in faeces
Another example where AI algorithms were used was on microbiome data from those suffering from inflammatory bowel disease (IBD). He says, ‘Often, those patients alternate between periods of inflammation and periods with no symptoms. We studied whether we could find strains that were linked to the inflammation period or even better, to the period where there are no symptoms. To do it we used a public data set containing information on people with IBD who for a year had their faeces tested for specific bacteria.’
Smart computer program
Mark: ‘By studying faeces, you can say something about the composition of bacteria in the intestines. In the dataset it was also specified when the people had symptoms and when they hadn’t. That was then input for AI algorithm. And indeed the outcome showed that certain bacteria are more abundant when inflammation arises. It also pointed towards certain bacteria that were linked to the period where there were no symptoms. That information can be used to look for new potential strains. That was of course a great start, but the data set contained just 132 people. For firm predictions you need data from a much bigger pool. Plus, we don’t know whether that bacteria was the cause or an effect of the inflammation.’
Just as with genome analysis, the lab research is indispensable to data technology. Mark explains, ‘The technology is advanced, but it’s not flawless. Furthermore, we don’t always have the vast amounts of the data we need at our disposal for reliable predictions. Until the computer models can offer us more certainty, the lab will always check the outcomes. But the predictions with the algorithms do provide an efficiency gain: we don’t have to research all our strains any more, just those that are positively identified with the algorithm.’
Some companies outsource their DNA analysis to external laboratories and merely use the results delivered to them. Mark says, ‘At Winclove Probiotics we have the expertise and the data technology in-house. That’s great, because we can compile more knowledge and understand the nuances of such analyses. And because we’re very well able to interpret the results ourselves, we then know the potential of our strains. That really accelerates and strengthens our innovation.’
Mark has set up the genome analysis process in R&D. He says, ‘That’s now achieved, and I’m proud of it. In the future I really want to do large-scale research into the microbiomes of the people who use our formulations. Then we can combine those results with an effect measurement for our formulations. Once you determine the relationship between the microbiome and the workings of our probiotics, you can make far more effective recommendations to improve the quality of life of our end-users. And that is something that we all want. ’