Authors:

Pranav Punuru

Date Created:

Not specified

Course Title:

Professor:

Not specified

About Paper:

The complexity and fragmentation of bioinformatics tools have long hindered efficient protein research, slowing down potential breakthroughs in structural biology and drug design. We introduce VIPER (Virtual Intelligence for Protein Exploration and Research), an AI-driven, conversational research assistant designed to revolutionize protein science by integrating and streamlining complex research processes. This study employs a novel integration of large language models (GPT-4), state-of-the-art protein structure prediction (ESMfold), and advanced molecular docking (DiffDock), all unified through an intuitive web interface built on FastAPI. VIPER demonstrates unprecedented capabilities in seamlessly connecting various bioinformatics tools, enabling complex workflows to be initiated with simple prompts. For instance, users can go from providing a protein name and a ligand to obtaining a 3D structure with the ligand docked, all in one single prompt. Moreover, VIPER can execute its own code and browse the web in real-time, significantly enhancing its ability to access and process the latest scientific information, run custom analyses, and adapt to new research challenges. This level of integration and adaptability streamlines research processes, potentially reducing the time and expertise required for complex protein analyses. The implications of VIPER extend beyond individual research tasks, potentially transforming entire drug discovery pipelines and enhancing our understanding of protein-ligand interactions on a broader scale. VIPER will be made available on ChatGPT.com, democratizing access to advanced protein research tools, and the code for the API will be provided, allowing for further development and integration into existing research workflows. Keywords: Bioinformatics; Protein Structure Modelling; Artificial Intelligence; Large Language Models; Research and Analysis

Source:

Purdue University / 2024

Topics:

No topics listed

Co-authors:

Pranav Punuru