Prediction and Analysis of Protein 3D Structures Using Protein Language Model and Streamlit

The development of deep learning has extensively amplified the accuracy of protein tertiary structure prediction. The success of AlphaFold from DeepMind demonstrates that extensive evolutionary information may be found in a sequence's multiple sequence alignment, which results in precise 3D models. It would take significant computing power to train a comparable model. Web apps can use strong server-side resources, so users don't always need high-end technology. This paper presents a novel web-based application that integrates ESMFold, a protein language model for protein structure prediction developed by Facebook AI, with Streamlit, an open-source webapp framework. Our application enables users to input an amino acid sequence and quickly receive a prediction of its tertiary structure. The analysis of the predicted structure is performed by Ramachandran Plot and calculating the pLDDT value. The effect of mutation on amino acid sequence is calculated by masked marginal score. Streamlit facilitates a user-friendly, interactive interface that requires minimal setup and can be used effectively by researchers without extensive computational expertise.