Proteins are the foundation of life, and designing functional proteins remains a key challenge in biotechnology. Before the development of AlphaFold2, the focus of design was primarily on structure-centric approaches such as using the well-known open-source software Rosetta3. Following the development of AlphaFold2, deep-learning techniques for protein design gained prominence. This study proposes a new method to generate functional proteins using the diffusion model and ESM2 protein language model. Diffusion models, which are widely used in image and natural language generation, are used here for protein design, facilitating the controlled generation of new sequences. The ESM2 model, trained on the basis of large-scale protein sequence data, provides a deep understanding of the context of the sequence, thus improving the model’s ability to generate biologically relevant proteins. In this study, we used the Protein A-like peptide as a model study object, combined the diffusion model and the ESM2 model to generate new peptide sequences from minimal input data, and verified their biological activities through experiments such as the BLI affinity test. In conclusion, we developed a new method for protein design that provides a novel strategy to meet the challenges of generic protein generation.

The virtual screening of synthetic proteins is to develop an effective method to predict and screen out synthetic proteins that are similar to the target sequences in terms of their structural and physicochemical properties by combining advanced computational tools and models, such as AlphaFold2 and ESM2. Through virtual screening, the experimental cost and time period can be reduced, and some guidelines for customized design of new proteins can be provided. All similar research efforts have promoted the advancement in the fields of synthetic biology and biomedical science. In this paper, we first calculate the backbone distances between synthetic proteins and target sequences using the AlphaFold2 tool to ensure that they are structurally similar. Secondly, based on the ESM2 model, the concept of feature distance is proposed to ensure that the two proteins are consistent in physicochemical properties. In the experiments, the steps of virtual screening of synthetic proteins are summarized, and the synthetic proteins are arranged in ascending order by backbone distance and feature distance. If the same synthetic protein appears in the first position of the two sorted tables, then select that protein. If different synthetic proteins appear in the first position of the two sorted tables, the synthetic protein with higher solubility will be selected. Repeat this process for other ranking positions. The experimental results show that this process derived from backbone distance and feature distance is a necessary and useful tool to select preferred synthetic proteins before entering the lab experimental session.

There are hundreds of millions of natural proteins in nature, and these natural proteins can play different roles due to their special structure. The first field of AI for Science is the prediction and design of protein structure, and simulating the structure of certain natural proteins has become a popular topic in biopharmaceutical research. The generative adversarial network model in deep learning and a target sequence method were used in the paper to mimic the special structure of natural proteins. We successfully synthesized the ProteinA-like peptide sequence which had a high structural similarity with the natural proteins in its tertiary structure.These synthetic peptide sequences by computer will lay the foundation for the later transcription to become mRNA and to synthesize protein in the laboratory .