Over the past year, scientists have renamed 27 human genes because Microsoft's Excel misread their names as dates and changed the formatting
Sometimes it's easier to rewrite genetics than update Excel — There are tens of thousands of genes in the human genome: minuscule twists of DNA …
The VergeJames Vincent
Context & Ripple Effects
The gene renamings were a workaround, not a fix: symbols like MARCH1 and SEPT1 were being silently converted into dates whenever researchers pasted them into Excel, corrupting published datasets. Rather than wait on a vendor, the scientific community changed its own nomenclature to avoid the collision.
Researchers and database curators who had adopted the 2020 renamings get a second migration path: with Excel no longer mangling gene symbols, legacy names like MARCH1 and SEPT1 become usable again, forcing choices about which nomenclature papers and pipelines standardize on.
Microsoft's own genomics-facing services — including Microsoft Genomics — inherit a credibility problem in a market where its flagship spreadsheet had corrupted the underlying data format.
Second-order effects
Bioinformatics tooling and journal submission systems that built workarounds around Excel's date conversion now face compatibility overhead in both directions, as old renamed datasets and restored legacy symbols coexist.
Competing analysis platforms can market fidelity to scientific nomenclature as a differentiator against Excel-based workflows, pressuring Microsoft to keep loosening defaults rather than reverting them.
Third-order effects
As genomics industrializes — AI models like DeepMind's AlphaGenome training on molecular data, cloud pipelines replacing desktop analysis — data provenance and format integrity shift from an inconvenience to a regulatory and reproducibility concern, with software vendors expected to accommodate scientific conventions rather than the reverse.
The pattern generalizes: general-purpose productivity software is being forced to add domain-specific escape hatches, and scientific communities are learning that naming standards are only as stable as the tools that read them.
The trend: Genomics is becoming a software-dependent discipline in which vendors, not scientists, increasingly absorb the cost of keeping machine-readable biology intact.
Find it fascinating that it's easier to rename human genes than to teach people how to disable Excel automatic date conversion https://www.theverge.com/... https://twitter.com/...
That Excel feature defeated quite a few bioinformatics papers as per this classic of 2004 https://bmcbioinformatics.biomedcentral .com/ ... and this one from 2016 https://genomebiology.biomedcentral.com / ... I wonder if anyone approached @MSFTResearch : Could heuristic AI veto n…
Am I getting this right? Microsoft created software that erroneously converted the names of genes into dates. This was such a problem for so long, that scientists officially changed the names of genes instead of Microsoft choosing to fix their software? https://twitter.com/...
Amazing that it's come to @genenames renaming gene symbols like SEPT1 and MARCH1 to avoid problems with Excel. Does this mean the end of my dream of giving some lncRNA of unknown function the gene symbol DROPTABLE? https://www.nature.com/...
Scientists are effectively doing SEO for naming human genes. OK, maybe it's EEO for Excel Engine Optimzation. Just another completely unexpected thing 2020 would like to offer up. https://twitter.com/...
Have you come across this issue in your genetic research? Sometimes it's easier to rename human genes than update Excel! https://twitter.com/... https://twitter.com/...
One study from 2016 examined genetic data shared alongside 3,597 published papers and found that roughly one-fifth had been affected by Excel errors. 🤦🏽 https://twitter.com/...
When we talk about technical debt, some decision made decades ago by a handful of people at Microsoft have impinged on genetic research: there's no method to ensure information won't be autoformatted in Excel. https://twitter.com/...