ADVANCING THE UNDERSTANDING OF INTRINSICALLY DISORDERED PROTEINS THROUGH STRUCTURAL ANALYSIS, EVALUATION OF PREDICTION TOOLS, AND DATABASE ENRICHMENT

This PhD thesis focuses on intrinsically disordered proteins (IDPs), a class of proteins that lack stable three-dimensional structures under physiological conditions but play crucial roles in numerous biological processes. It centers on developing computational tools, methodologies, and resources to analyze protein structures and predict intrinsic disorder from primary sequences. The first component of this study is DRMAAtic, a tool designed for efficient distributed resource management in an HPC environment. By integrating APIs and web server functionalities, it enables large-scale exploitation of computational clusters, essential for processing high-throughput analyses, providing different tools for managing resources, users, and a fair use of the system. Next, the Residue Interaction Network Generator (RING) is introduced, a tool that analyzes and visualizes residue interactions in protein structures, including various interaction types and contact distributions. This is critical for understanding protein dynamics and interactions. Coupled with DRMAAtic, RING forms the basis of a powerful web platform for the scientific community, offering advanced tools for the analysis of protein structures and interactions. Additionally, the thesis presents the methods and results of the second round of the Critical Assessment of Intrinsic Disorder Prediction (CAID) experiment, which evaluated the performance of state-of-the-art disorder predictors. From this, the CAID Prediction Portal was developed using DRMAAtic as its foundation, integrating multiple predictors into a unified web platform. This portal provides a valuable resource for accurate and comprehensive predictions of IDP regions and binding sites. Lastly, the thesis explores the progress made in enhancing MobiDB, a comprehensive database for protein disorder and mobility annotations. MobiDB consolidates information from literature, experimental evidence, and predictions, providing a unified perspective on the disorder landscape across all known protein sequences. The discussion will cover recent advancements in MobiDB, including the incorporation of new data sources, upgraded APIs, and improved visualization tools. This study advances computational biology by providing essential tools and resources for analyzing protein structures and interactions, predicting intrinsic disorder, and providing a core resource for the scientific community such as MobiDB. The evaluation of disorder predictors and the development of the CAID Prediction Portal set new standards for the field, raising the bar for the accuracy and reliability of disorder predictions.