Why Mouse Models in Biomedical Research Do Not Always Translate to Humans

Imagine two identical-looking smartphones. They have the same outer casing, but one runs iOS and the other runs Android. They look identical, but they behave completely differently when you tap an app.

This is the central challenge when using mouse models in biomedical research. We share about 85 percent of our protein-coding DNA with mice, yet therapies that cure diseases in rodents frequently fail in human trials. For decades, scientists assumed high DNA similarity meant high functional similarity. This assumption often leads to costly dead ends in clinical trials because the way cells actually read those blueprints—via RNA—tells a different story.

Upgrading Mouse Models in Biomedical Research

Researchers used an AI model called GeneRAIN to analyse RNA data from 777,000 human and mouse tissue samples. The system operates like a translation dictionary, converting raw genetic data into functional profiles. Instead of just reading the static DNA code, they measured how genes actually behave in action.

The study identified key differences:

• Hidden Discrepancies: 2,407 genes have nearly identical DNA sequences but show completely different RNA behaviours, meaning they likely link to different diseases.
• High-Risk Targets: 3,070 genes differ at both the DNA and RNA levels, presenting a high risk of laboratory discrepancies.
• New Links: The AI successfully mapped elusive "long noncoding RNA" matches across both species.

This map suggests why some therapies fail to cross the species gap. It could allow scientists to predict when a mouse model is likely to mislead them, helping to organise safer, more accurate clinical trials.