Discrepancy in GEO2R vs. Raw Data Analysis: Significance of DEGs

Question

[Saeedeh Salehi]

Hello everyone,

I used the GEO2R tool for differential expression analysis (DEA) and obtained differentially expressed genes (DEGs) between two conditions. After downloading the results table from GEO, I noticed that for my target gene, there were multiple probe IDs, each showing different logFC and adjusted p-values—some significant and some not.

However, when I downloaded the raw data and performed preprocessing (including aggregating duplicate genes), the adjusted p-values for my target gene were no longer significant. My Questions:

1. Does this discrepancy suggest that GEO2R is not a reliable tool for DEA?
2. If only one or two probe IDs for a gene show a significant adjusted p-value, can the gene still be considered significantly differentially expressed?
3. Should raw data analysis always be preferred over using GEO2R, especially when dealing with multiple probe IDs per gene?

I appreciate any insights or recommendations on best practices!

Thanks in advance.

Answer 1

[ATpoint]

GEO2R is very primitive. I would always download raw data and process myself. That is more reproducible. Then results are what they are, given analysis was not done incorrectly. You might want to take the probe with highest expression per gene rather than collapsing somehow. This was asked many times before, please search "multiple probes per gene". Essentially, it's capturing isoforms.