Yangwenshan Ou conducts her PhD research in the Molecular Ecology group at Microbiology Laboratory, Wageningen University and Research (WUR), The Netherlands. She specializes in the development of gut microbiota from birth to puberty and the link between behavior and gut microbial compositions in children.
“Can disturbances in the gut microbiota composition cause behavioral problems?” That is the question frequently asked when I present my studies. In the past several years, I have been looking into the correlations of the gut microbiota composition with problem behavior in healthy children. But correlation is not the same as causation. In this blog, I will explain what is needed to answer the above question about causation.
First, to explain the difference between correlation and causation, think about this example: on a warm summer day, there is a high correlation between the number of ice creams sold, and the number of people that drown. So, do ice creams cause drowning? That is unlikely. Ice cream sales and drowning are highly correlated, but not causally linked (this example is explained in more detail by professor Ionica Smeets in a Ted X talk).
So, back to my own research. I want to know if disturbances in the gut microbiota composition cause behavioral problems. Unlike animal studies in which researchers can investigate causations in a much more feasible way, human studies sometimes are hampered by the difficulties in answering causations. One option is to set up an interventional study with human participants, for instance where one group receives a probiotic and the other group does not. The problem is that many of these studies are still very exploratory and are not yet probably testing any concrete hypotheses.
Are there any strategies that we may use for the aim of causations? I will share some rethinks based on my prior experiences and hope this would inspire other researchers with similar confusions. Note that the following text may get a bit technical for non-researchers.
1. Longitudinal research
To get closer to the causations, I would recommend using a longitudinal study rather than a cross-sectional one. Longitudinal studies allow repeated measures in a community sample over a certain period, and this ensures better robustness of results. For instance, if we find the abundance of a gut bacterium (i.e., how much it is present in the gut) correlates with problem behavior at multiple time points, then we can be more confident to assume a potential causation.
2. Dealing with missing data
What to keep in mind is that analyzing data from a longitudinal study is not an easy task. One of the challenges is to handle missing values. This can be a big issue for longitudinal studies comprising a small sample size and lasting over the years. In this case, imputing missing values can be a good approach (this means that you make an educated guess about what the missing values are, based on the data that you do have). Fortunately, there are many available tools for data imputation, like the mice package for R.
3. Reduce Bias
Another challenge of longitudinal studies is to account for covariates (i.e., variables that influence both predictors and outcomes) to reduce the bias. To date, there are some covariates, like age and diet, known for their impacts on the gut microbiota composition and mental health, and hence it is necessary to consider them when analyzing data.
Since we have moved so far, what is the next step after finding a robust correlation between bacterial abundance and problem behavior across time? There are some rethinks:
4. Beyond 16S: Measure the precise taxonomic level of bacteria
To identify which bacteria are present in the gut, we most of the time use 16S rRNA gene sequencing. This is a quick and affordable technique that gives us a lot of information about the bacteria that are present. However, no technique is flawless. 16S rRNA gene sequencing shows low resolution at the species level and is unable to discriminate some genera. Therefore, it is necessary to quantify target bacteria at a more precise taxonomic level. Quantitative PCR is a technique widely used in molecular labs. It can be used to detect, characterize and quantify specific bacteria.
5. Measure the metabolites that the bacteria produce
Researchers should also check whether target bacteria hold specific genes in producing mental health-related metabolites, such as short-chain fatty acids (SCFAs), γ-aminobutyric, and serotonin. An increasing number of animal studies have shown that the gut microbiota can send a molecular signal to the brain and then affect behavior. To further explore this idea, measuring fecal concentrations of these metabolites would help.
6. Isolate and cultivate bacteria
Sometimes, the bacteria we find with fascinating correlations are only detected in sequencing and never be cultivated. This makes bacterial isolation and cultivation a necessity, especially when we want to use the target bacteria as a future intervention, for instance in a probiotic treatment.
In a nutshell, causations are what we want to answer in the end, while associations are what we need to figure out in the first step. With more and more research being done on the gut microbiota, it is time for researchers to take the next steps towards identifying causations.