Software is subbing mistakes into scientific papers

MICROSOFT EXCEL’S zeal for autocorrection has long irritated casual users. Lists of numbers preceded by dashes are misread as instructions to subtract. Phone numbers lose their leading zeroes. Credit-card numbers get re-expressed in scientific notation.

Geneticists struggle with a particular version of this problem. A gene called Membrane Associated Ring-CH-type finger 1, commonly known as MARCH1, is, for instance, frequently re-encoded as the date March 1. Something similar happens to genes in the Septin family, of which SEPT1 is a member, and to Basic Helix-Loop-Helix Family Member E41, often known as DEC2.

This problem was first noticed in 2004, but was brought to wider attention in 2016 by Mark Ziemann of Deakin University, in Australia. In July Dr Ziemann followed up with a paper in PLOS Computational Biology entitled “Gene name errors: Lessons not learned”. By surveying 166,000 genomics-related papers published between 2014 and 2020, he and his co-authors showed that the number of papers using Excel has steadily increased, and the proportion plagued with autocorrect errors continues to hover at around 30%.

Errors have also been flagged by researchers in other languages. In Portuguese, for instance, AGO2 (Argonaute RISC ...