How to address conflicting circRNA expression results between public datasets and published studies? is it normal?

Question

[ParastooA]

Hi all, I’m working on circRNA expression analysis in colorectal cancer and encountered a common but confusing issue. A few circRNAs I found to be differentially expressed in a public dataset show the opposite expression pattern in some published experimental studies (e.g., upregulated in papers, downregulated in the dataset).

I understand this may stem from differences in platforms (e.g., microarray vs RT-qPCR), tumor stage, or cohort variability. Is it normal at all? I am doubting all my data.

My questions:

1.  How should I acknowledge and discuss this discrepancy in a paper or thesis?
2.  Is it acceptable to continue highlighting such circRNAs as potential biomarkers despite inconsistency?
3.  Have you seen similar situations? How do researchers typically approach this?

Appreciate your input!

[rfran010]

1. How should I acknowledge and discuss this discrepancy in a paper or thesis?

The differences you mention is a good start. Would need more careful, critical analysis and comparison of the two papers. Is there a technical difference that could explain, or a biological? Are other factors in agreement between the papers? Does one paper have more robust, consistent, and orthogonal data to support it?

1. Is it acceptable to continue highlighting such circRNAs as potential biomarkers despite inconsistency?

Potentially, if one paper seems more robust. But, in general something should not be considered a "biomarker" without various validations. I may describe it as a potential biomarker based on your description though, of course dependent on what a careful analysis might reveal.

1. Have you seen similar situations? How do researchers typically approach this?

I think there's a fair amount of variability in data when it comes down to specifics. I think that's the nature of biological (and others) science, due to various biological and technical differences. If overall trends are inconsistent it's more of a cause for concern. When researching a specific gene, though, I wouldn't be surprised to see different behaviors especially if there are significant methodological differences. I think the best way to approach this is to critically evaluate the evidence and see if one has more evidential weight. If both are equal, or equally weak, then the data should be taken with a grain of salt.

Generally, when discussed in a research paper, it is concise: "So and So showed it was up while They showed it is down. However, They cultured in conditioned media while So and So used defined media. Conditioned media has been shown to affect the Gene Pathway and may explain this discrepancy" or "There are some discrepancies in whether This is a biomarker. Methodology has widely varied in these studies so more careful, focused analysis is needed."

In a thesis, I would add more nuanced or detailed commentary if it's a significant part of the study, while in a paper the concision is preferred.