Thermophilic proteins are known for their stability in higher temperatures, which makes them ideal for industrial applications. However, their use is limited due to their lower activity levels than their mesophilic counterparts. In this project, we propose a novel approach to transfer the thermal stability property of thermophilic proteins to mesophilic proteins using a GAN. Our method also uses LLMs such as ESM-1b to keep the proteins syntactically correct, which preserves their function.
The novelty in your approach is the ability to efficiently train the model to generate proteins with specific properties, such as thermostability while leveraging the power of LLMs such as ESM-1b.