* This dataset supports online use.Click here to jump.

VenusMutHub is a new benchmark for evaluating protein mutation engineering AI models released by Professor Hong Liang's team at Shanghai Jiao Tong University in 2025. The related paper results are "VenusMutHub: A systematic evaluation of protein mutation effect predictors on small-scale experimental data", the study has been published in the journal Acta Pharmaceutica Sinica B.

VenusMutHub is the first small sample dataset of protein mutations for real application scenarios. The research team carefully compiled 905 small sample experimental mutation datasets for real application scenarios, covering 527 proteins (of which 98% has 5-200 mutations), and covering various functional measurement data such as stability, activity, binding affinity and selectivity. All data are measured using direct biochemical measurements rather than alternative fluorescence readings to ensure the accuracy of the evaluation.

In addition, the team also tested 23 cutting-edge computational models (AI and non-AI models), covering a variety of methods that utilize protein sequences, three-dimensional structures, and evolutionary information, and comprehensively evaluated the predictive capabilities of these models on small sample data in real application scenarios.

As an open resource, VenusMutHub will continue to be updated (https://lianglab.sjtu.edu.cn/muthub/), providing a reliable benchmark for the field and promoting protein function prediction towards more standardized and faster development.