MobiDB: 10 years of intrinsically disordered proteins

Abstract

The MobiDB database (URL: https://mobidb.org/) is a knowledge base of intrinsically disordered proteins. MobiDB aggregates disorder annotations derived from the literature and from experimental evidence along with predictions for all known protein sequences. MobiDB generates new knowledge and captures the functional significance of disordered regions by processing and combining complementary sources of information. Since its first release 10 years ago, the MobiDB database has evolved in order to improve the quality and coverage of protein disorder annotations and its accessibility. MobiDB has now reached its maturity in terms of data standardization and visualization. Here, we present a new release which focuses on the optimization of user experience and database content. The major advances compared to the previous version are the integration of AlphaFoldDB predictions and the re-implementation of the homology transfer pipeline, which expands manually curated annotations by two orders of magnitude. Finally, the entry page has been restyled in order to provide an overview of the available annotations along with two separate views that highlight structural disorder evidence and functions associated with different binding modes.

Insights into the conservation and diversification of the molecular functions of YTHDF proteins

“…We next focused on the N-terminal regions of ECT1/9/11 . Because protein disorder predictions using MobiDB (66) confirmed that they are all disordered (Fig. S11), we used approaches suitable for the study of IDRs.…”

The 2023 Nucleic Acids Research Database Issue and the online molecular biology database collection

“…The update from GPCRdb ( 27 ) reports state-specific AF2 models alongside other new features such as lists of ligands for each receptor, both endogenous and surrogate. Even MobiDB ( 28 ), focusing on intrinsically disordered proteins, benefits from two AF2-related predictions unforeseen by the original methods developers: predictions of disordered regions and potential interaction motifs contained within them. Other classes of proteins with special structural properties are covered by the new Amylograph ( 29 ) which curates information on amyloid-amyloid interactions and the returning PhaSepDB ( 30 ) for proteins that can participate in phase separation, now doubled in size and with much more detailed annotations.…”

Biophysical and Integrative Characterization of Protein Intrinsic Disorder as a Prime Target for Drug Discovery

“…Several machine learning methods, from simple linear regression to more sophisticated approaches such as support vector machines and deep learning (artificial neural networks with multiple layers), can combine existing sequence descriptors to predict the disordered sequences. With the increasing degrees of freedom (sequence descriptors) and increasing training datasets (e.g., DisProt [ 108 ], IDEAL [ 109 ], MobiDB [ 110 ], and solved PDB structures), deep learning has become a commonly used method for this purpose. A recent assessment testing 43 predictors found that machine learning methods and specifically deep learning methods outperform physicochemical methods [ 111 ].…”