Do missed patterns warrant correction to a published paper?

Question

hoping some scientist academics out there can provide insight into a question I've been ruminating on for a while.

I spent multiple years working on a large computational data set, and it was published in a journal this year. There were millions genes and many interesting patterns in this project. I am now working on another large data set, collected with different organisms, and going back to my original code to optimize and try out new methods for data exploration. In doing this, I discovered better ways of carrying out my original analysis. I wanted to understand how these modifications changed the biological outcomes in my already published study. In doing so, I discovered a pretty interesting gene that I missed the first time around, that was overall not very abundant in the scheme of things, but definitely responsive. It's a part of a module that I discussed during a section of this paper. It doesn't change major conclusions, but if I could, I would rewrite a part of the paper to take into account this missing gene. The concern is that I will lead people astray who have a very specific interest in this gene process.

My question is, how should one handle this? Do I correct the paper? If so, do I merely mention this gene I missed or re-do the entire analysis using an updated (and better) pipeline? I have not seen this done in practice, and the corrections/errata I see are due to technical and specific errors, not to add in extra information. It seems like a Pandora's box because I now know so much more than I did the first time I did the analysis, and could probably keep updating every time I find another new important gene that was previously missed or a better way of carrying out the analysis. Another option is to write a new paper expanding on this subsection.

I can't tell at what point I'm crossing the line from being a diligent scientist to obsessing over every detail. I know things like this must happen to other people, but I don't hear of them being discussed. Is anyone else concerned about how scientific methods (especially computational ones) and our own data analysis skills improve over time, and how this influences previous results? How do others deal with this in academia?

Answer 1

What you are describing does not sound like a correction to me --- it sounds more like something where a follow-up paper would be warranted. There are many cases where an initial paper is published with an analysis of data, and then later follow-up papers use different models, account for additional variables, etc. If you think that the new analysis you are proposing is better, and adds to knowledge on the topic, I would say that this is a good reason to write a short follow-up paper with your new analysis.

Answer 2

If I understood you correctly, what you are describing is really a new research result. You designed a new method, and using that new method on old data you discovered a new result. In retrospect you realize that you could have seen a hint of the new result with your old methods, but it was not obvious without your new way of approaching the problem.

A correction should generally be for an error that invalidates some result or conclusion. Here is a definition from the Physical Review style guide (this is a common journal from my field -- I realize this is isn't directly relevant for genetics, but I think the principle should be universal and it gives you an idea of what to look for in the journal you are interested in publishing in):

Errata: The Errata section contains notices regarding errors or omissions in papers previously published. Besides standard Errata, other categories of documents may appear in this section. Each has bidirectional links between the original article and the document in the Errata section. The category of the corrective document is indicated in its title and in the link from the original article. The standard Erratum is a statement by the authors of the original paper that briefly describes the correction(s) and, where appropriate, any effects on the conclusions of the paper.

In this context I would interpret "omission" to mean "failing to mention an important piece of context" (eg: a reference, a crucial detail in the methods, a counterexample to a trend reported in the paper), and not new research on old data. There are several major disadvantages to publishing a correction, rather than a new paper: (a) most people probably won't notice the correction (far fewer people compared to a new paper), (b) the people who do notice will probably assume it was a mistake (which is neutral or negative), as opposed to what really happened, which is a new discovery enabled by improved methods (which is a clear positive).

If anything I think you should actually be excited about your result and want to tell people! I can think of two common approaches to this kind of situation:

1. If the result by itself is interesting enough, you could write a follow-up paper which describes the new result.

2. Otherwise, as part of a followup paper describing the new method applied to the new data set, you can have a section entitled something like "re-analysis of old data set" where you show how your new method enables a new insight.

In either case, I would suggest you describe both how you discovered the new insight with the new method, and how you could have seen hints of the result with the old methods (but it would have been harder). People like stories, and it sounds like this will tell the story of an interesting result and the power of your new method.

Answer 3

When you find new information which adds a small amount to work that is already published, that new information should be published in a "comment" or "matters arising." Unfortunately, some journals do not publish such comments or matters arising.

If you use arXiv you can just update your preprint.