A-DOBRA: HIGH PERFORMANCE WEBSERVICE APPLIED TO STRUCTURAL BIOINFORMATICS
HPC; Protein modeling; Workload; Containers; Webservices
Predicting the three-dimensional structure of proteins represents one of the greatest challenges in modern molecular biology, being fundamental for the development of multiple areas in society. Experimental methods such as X-ray crystallography, nuclear magnetic resonance, and electron cryomicroscopy have long been employed to determine these structures; however, they require considerable resources and extensive time, in addition to not being applicable to all proteins. This limitation has driven the development of computational approaches based on artificial intelligence, particularly AlphaFold, a technology developed by DeepMind that revolutionized structural prediction through attention neural networks and training with evolutionary databases. Therefore, the objective of this study is to develop A-DOBRA, a high-performance webservice integrated with high-performance computing (HPC) systems to democratize access to protein modeling using AlphaFold, providing an intuitive platform for natural science professionals without computational expertise. The methodology involved significant optimizations in the AlphaFold 2 pipeline, including improvements to download scripts for genetic and structural databases, implementation of integrity checks, and alternative pathways for error handling. However, with the release of AlphaFold version 3, the approach had to be updated to enhance modeling results with the most modern tools available in this scientific field. Apptainer containers were used for environment isolation and integration with SLURM workload manager for computational resource optimization. In web development, PHP programming language was used with the Laravel framework and PostgreSQL database.