Motivation

It was all over the news; Organic apples contained a healthier microbiota community compared to conventional grown apples. Eat the whole apple to get 100 million bacteria and eat an organic apple to avoid potential food-borne pathogens. Headlines based on a scientific study by Wassermann et al. (2019). I think it is a good habit not to believe everything you read, but the story caught my attention. I like apples, I believe in organic farming, and I was on the lookout for a more recent dataset for a microbiota workshop. I needed a 16S amplicon based dataset small and simple in design. This dataset seemed perfect.

While I was waiting for the release of the raw data, I gave the paper a more critical reading. Instead of clarity, I got questions. A lot of questions. The experimental design seems clear, but I struggled to understand even the most basic box plots. How is it possible to have over 4 outliers while we only have 4 data points? The high number of OTUs stunned me too. It is possible to have 92,365 OTUs in and on apples? I did not understand why the authors would rarefy the OTUs down to 1,525 and lose ~ 98% of the data in doing so. The comparison between the treatments seemed not fair. The non-organic apples were not fresh from the tree but bought from a supermarket. Where are the food-born pathogens located? The focus was on rare taxa with no negative controls. I am not unfamiliar with microbiota data analysis, but I could not understand what the authors did. The description of the material and methods fluctuated from meticulous to obscure and sometimes even contradicting. I had a strong feeling that the authors wanted to prove a point rather than testing a hypothesis. Some obvious mistakes might also shine a poor light on the devotion of the reviews.

My first reaction after having read the paper more carefully was not to use this study for any student course. I changed my mind. I think it is an excellent example to teach students about microbiota data analysis, data interpretation, and reproducibility. 

I do not intend to discredit the authors. We all make mistakes. It is also not my intention to blame the authors for a simplified media coverage where ignorance can provide headlines. I do this because I need to understand the data analysis to believe the results and trust the interpretation of the data. I am thankful to the authors for the paper and the dataset. 

Write Me

This is work in progress. I will do my best to explain my findings and compare it with the paper. Please contact me if you have something to contribute to the discussion (jwalser[at]ethz.ch). No worries, I will not include anybody's contribution(s) without consent.