# Deep learning in biology faces a transferability crisis

## TL;DR

The paper discusses how misleading claims about model transferability in deep learning are causing problems in biology, and suggests ways to fix it.

## Contribution

The paper highlights the transferability crisis in deep learning for biosciences and proposes solutions to improve model evaluation.

## Key findings

- Misleading claims about model transferability are undermining reliable performance evaluation in biosciences.
- The paper emphasizes the need for better practices to ensure generalizable deep learning models in biology.

## Abstract

Creating generalizable models is a conserved aim in deep learning—however, misleading claims of transferability threaten to obfuscate reliable performance evaluation. We outline the severity of this issue in the biosciences, and suggest potential solutions.

Creating generalizable models is a conserved aim in deep learning—however, misleading claims of transferability threaten to obfuscate reliable performance evaluation. This Perspective article outlines the severity of this issue in the biosciences, and suggests potential solutions.