Making AlphaFold2 accessible to Biologists

AlphaFold2 has revolutionized medical research with its unparalleled accuracy in protein structure prediction, vital for understanding biological functions and advancing therapeutic developments. However, its deployment poses significant challenges due to its complexity and the high computational resources required. Many biologists often lack the technical skills to run such advanced models efficiently. In this presentation, we will address these challenges, and how we restructured the AlphaFold2 process into three streamlined stages: feature extraction and multiple sequence alignment (MSA), inference (structure prediction), and relaxation, with an additional module for plotting results. The feature extraction and MSA stage requires CPUs only to run, while the inference stage utilizes GPUs to predict the protein s three-dimensional structure, maximizing computational efficiency. The relaxation stage refines these structures into more biologically plausible models. This segmentation allows for separate processing, conserving GPU resources for the computationally intensive inference stage and running other stages on CPUs. The next challenge was making AlphaFold2 easy for biologists with minimal experience in HPC to launch. Using Nextflow and Nextflow Tower (Seqera Platform) have been crucial in overcoming this challenge. Nextflow encapsulates and manages complex workflows, simplifying the multi-step process into manageable parts. Nextflow Tower offers a graphical interface that enhances the user experience making it easy to launch, monitor, and manage their workflows. These advancements ensure that resources are used efficiently and that biologists with limited technical skills can leverage the latest protein modeling technology, significantly enhancing research capabilities.