On July 2, 2022, we retracted a paper we published last year in Evolution https://onlinelibrary.wiley.com/doi/10.1111/evo.14551. The reason I wanted to write this post is to explain what happened and how we dealt with it and thereby to help normalize honest retractions which should probably be more common. I’ll give a little detail on the experiment and then focus on the human side.

The experiment was an exciting one. We took several clones of the social amoeba we study, Dictyostelium discoideum, cured them of the bacteria that they carried symbiotically, and then let them proliferate through many generations isolated from their bacteria. We also took those same bacteria and let them proliferate on their own for many generations. If you evolve to make your partner worse when reintroduced, you would originally have been adapted to cooperate with them. And if you evolve to harm your partner less when reintroduced, you would originally have been adapted to exploit them. We feel it is a new approach using lab evolution to understand what originally happened in the wild.

We got a clear experimental response and published the paper. It was only later when we decided to sequence the lines to see what changed that we discovered the problem. Sequences from lines that should have been Paraburkholderia. hayleyella turned out to match P. agricolaris. There had been cross contamination. This was not a rare event but impacted all the P. hayleyella lines. By the time we did the final experiment mixing host and bacteria, all the bacterial lines were predominantly P. agricolaris...

...Retraction has a stigma about it... ...When I discovered the contamination, I could have quietly moved on and likely nobody would have ever known. Some selfish, anxious part of me wanted to do that. But I believe in the importance of intellectual honesty and owning my mistakes...

...The first few results that did not turn out as I expected I assumed were because I had made some mistake in my code, had failed to set some needed argument, or failed to understand one of the multitude of assumptions inherent to doing bioinformatics. Each would send me off on some tangent trying to understand some new aspect of the program I was using.

Eventually, though, the simpler, uglier explanation occurred to me: I had messed up. I had messed up bigtime. Not with a few lines of code that could be fixed with some careful Google research. I had messed up the experiment itself, many months and dollars ago...