From Sequence to Structure: A Web-Based Gateway to AlphaFold Through Ohio Supercomputer Center Resources
Das, Pranav; Duan, Williams; Mishra, Shaurya; Wang, Robin; Tang, Joseph
- Publisher
- Zenodo
- Language
- en
Abstract
Obtaining experimental protein structures is time- and resource-intensive, often requiring highly trained scientists and months or even years of work. AlphaFold has emerged as a breakthrough AI/ML tool, capable of predicting protein structures at atomic resolution without relying on known structural templates. This represents a powerful example of how artificial intelligence can accelerate discovery in biology and medicine. However, many potential users of AlphaFold lack access to high-performance computing (HPC) environments or familiarity with command-line interfaces (CLI), Linux terminals, and shell scripting. To address this barrier, we developed SigmaFold (http://GitHub.com/Lumonike/SigmaFold), a web-based application that enables streamlined use of AlphaFold2 through an intuitive graphical interface. SigmaFold simplifies the workflow by integrating directly with the Protein Data Bank (PDB): users can input a PDB ID to automatically retrieve amino acid sequences, generate FASTA files, and launch AlphaFold predictions on the OSC’s HPC resources. The application also automates multiple sequence alignment (MSA) generation and renders 3D structures using 3Dmol.js. It presents pLDDT (predicted local distance difference test) and PAE (predicted aligned error) accuracy plots in real time, eliminating the need for external visualization tools or custom scripts. While AlphaFold3 now offers a public GUI (https://alphafoldserver.com/), AlphaFold3 itself is still a “black-box” experience. In contrast, our application was purpose-built for education and research training. It emphasizes transparency, reproducibility, and full workflow understanding, offering students direct engagement with the underlying biology, computation, and AI concepts. Our work supports meaningful participation from learners of all skill levels and provides a hands-on, HPC-integrated gateway into the world of AI-powered protein prediction. Acknowledgement: This project was developed as part of the STEM Institute 2025 with generous support and high-performance computing resources provided by the Ohio Supercomputer Center, a member of the Ohio Technology Consortium.
Full text
From Sequence to Structure: A Web-Based Gateway to AlphaFold Through Ohio Supercomputer Center Resources Pranav Das, Williams Duan, Shaurya Mishra, Robin Wang, Joseph Tang Ohio Supercomputer Center (OSC), 1224 Kinnear Road, Columbus, OH, 43212, US Key words: AlphaFold, WebApp, Protein Folding, Computational Biology Obtaining experimental protein structures is timeand resource-intensive, often requiring highly trained scientists and months or even years of work. AlphaFold has emerged as a breakthrough AI/ML tool, capable of predicting protein structures at atomic resolution without relying on known structural templates. This represents a powerful example of how artificial intelligence can accelerate discovery in biology and medicine. However, many potential users of AlphaFold lack access to high-performance computing (HPC) environments or familiarity with command-line interfaces (CLI), Linux terminals, and shell scripting. To address this barrier, we developed SigmaFold (http://GitHub.com/Lumonike/SigmaFold), a web-based application that enables streamlined use of AlphaFold2 through an intuitive graphical interface. SigmaFold simplifies the workflow by integrating directly with the Protein Data Bank (PDB): users can input a PDB ID to automatically retrieve amino acid sequences, generate FASTA files, and launch AlphaFold predictions on the OSC’s HPC resources. The application also automates multiple sequence alignment (MSA) generation and renders 3D structures using 3Dmol.js. It presents pLDDT (predicted local distance difference test) and PAE (predicted aligned error) accuracy plots in real time, eliminating the need for external visualization tools or custom scripts. While AlphaFold3 now offers a public GUI (https://alphafoldserver.com/), AlphaFold3 itself is still a “black-box” experience. In contrast, our application was purpose-built for education and research training. It emphasizes transparency, reproducibility, and full workflow understanding, offering students direct engagement with the underlying biology, computation, and AI concepts. Our work supports meaningful participation from learners of all skill levels and provides a handson, HPC-integrated gateway into the world of AI-powered protein prediction. Acknowledgement: This project was developed as part of the STEM Institute 2025 with generous support and high-performance computing resources provided by the Ohio Supercomputer Center, a member of the Ohio Technology Consortium.
