Unraveling causality in microbiome research
By eating mozzarella, you can contribute to the number of civil engineering doctorates awarded. Don’t believe us? Just look at the following graph:
It is even statistically significant, yielding a p-value of 0.000012. But before you go emptying out your local store’s mozzarella supply, you might want to take this with a grain of salt: correlation does not imply causation. While this peculiar relation between these two facts could mean that one causes the other, it is a lot more plausible that these either have a common cause or just happen to coincide in this particular timeframe.
This holds especially true in our microbiome, the collection of bacteria, viruses, fungi and other microbes that inhabit our bodies. Concurrent trends have led researchers to establish associations between bacterial populations in our gut and obesity, Alzheimer’s, COVID19, and many more health issues. From a pharmaceutical perspective, these associations are extremely interesting. Since we can modulate our microbiome through nutrition, pro- and antibiotics, we should hypothetically be able to influence the development of these diseases. All of this has driven hundreds of millions of USD into drug development of microbiome-based therapeutics.
While you were probably skeptical enough to spot the leap in logic in the mozzarella example, untangling the web of correlations in biological data is often a lot harder to do as there are more factors at play.
However, testing every association in the lab takes time and can even be unethical, since we can’t just give people a disease to see how they respond. In the context of a VLAIO research grant, BioLizard is now investigating the causality behind microbiome-disease associations. Taking into account the broadest picture of microbes, patient background and interactions, we are building a framework using state-of-the-art AI tools to evaluate causal links in the human microbiome, purely from data. This way we can check whether microbe A causes disease B or if they’re only distantly related, refining our focus. Our goal with this research is to support biopharmaceuticals by discovering drug targets and biomarkers supported by real cause-effect relations.