Computational Methods to Predict Intrinsically Disordered Regions and Functional Regions in Them.

Intrinsically disordered regions (IDRs) are protein regions that do not adopt fixed tertiary structures. Since these regions lack ordered three-dimensional structures, they should be excluded from the target portions of homology modeling. IDRs can be predicted from the amino acid sequences, because their amino acid compositions are different from that of the structured domains. This chapter provides a review of the prediction methods of IDRs and a case study of IDR prediction.

How AlphaFold2 Predicts Conditionally Folding Regions Annotated in an Intrinsically Disordered Protein Database, IDEAL.

AlphaFold2 (AF2) is a protein structure prediction program which provides accurate models. In addition to predicting structural domains, AF2 assigns intrinsically disordered regions (IDRs) by identifying regions with low prediction reliability (pLDDT). Some regions in IDRs undergo disorder-to-order transition upon binding the interaction partner. Here we assessed model structures of AF2 based on the annotations in IDEAL, in which segments with disorder-to-order transition have been collected as Protean Segments (ProSs). We non-redundantly selected ProSs from IDEAL and classified them based on the root mean square deviation to the corresponding region of AF2 models. Statistical analysis identified 11 structural and sequential features, possibly contributing toward the prediction of ProS structures. These features were categorized into two groups: one that contained pLDDT and the other that contained normalized radius of gyration. The typical ProS structures in the former group comprise a long α helix or a whole or part of the structural domain and those in the latter group comprise a short α helix with terminal loops.

Exon Elongation Added Intrinsically Disordered Regions to the Encoded Proteins and Facilitated the Emergence of the Last Eukaryotic Common Ancestor.

Most prokaryotic proteins consist of a single structural domain (SD) with little intrinsically disordered regions (IDRs) that by themselves do not adopt stable structures, whereas the typical eukaryotic protein comprises multiple SDs and IDRs. How eukaryotic proteins evolved to differ from prokaryotic proteins has not been fully elucidated. Here, we found that the longer the internal exons are, the more frequently they encode IDRs in eight eukaryotes including vertebrates, invertebrates, a fungus, and plants. Based on this observation, we propose the "small bang" model from the proteomic viewpoint: the protoeukaryotic genes had no introns and mostly encoded one SD each, but a majority of them were subsequently divided into multiple exons (step 1). Many exons unconstrained by SDs elongated to encode IDRs (step 2). The elongated exons encoding IDRs frequently facilitated the acquisition of multiple SDs to make the last common ancestor of eukaryotes (step 3). One prediction of the model is that long internal exons are mostly unconstrained exons. Analytical results of the eight eukaryotes are consistent with this prediction. In support of the model, we identified cases of internal exons that elongated after the rat-mouse divergence and discovered that the expanded sections are mostly in unconstrained exons and preferentially encode IDRs. The model also predicts that SDs followed by long internal exons tend to have other SDs downstream. This prediction was also verified in all the eukaryotic species analyzed. Our model accounts for the dichotomy between prokaryotic and eukaryotic proteins and proposes a selective advantage conferred by IDRs.

Classification of proteins inducing liquid-liquid phase separation: sequential, structural and functional characterization.

Liquid-liquid phase separation (LLPS) within the cell can form biological condensates, which are increasingly recognized to play important roles in various biological processes. Most proteins involved in LLPS are known to be intrinsically disordered proteins containing intrinsically disordered regions (IDRs) with low complexity regions (LCRs). The proteins driving LLPS were selected from databases of LLPS-related proteins and then classified into three classes according to the components in the condensates. Through in silico analyses, we found that proteins in the homo class, those that induce LLPS without partner molecules, have different IDRs and LCRs compared with the reference proteome. In contrast, proteins in the other classes, those that induce LLPS with partner proteins (the hetero class) or nucleic acids (the mixed class), did not show a clear difference to the reference proteome in IDRs and LCRs. The hetero-class proteins contained structural domains to serve protein-protein interactions, and the mixed-class ones had the structural domains associated with nucleic acids. These results suggest that IDRs in the homo-class proteins have unique IDRs, which provide multivalent interaction sites required for LLPS, whereas the hetero- and mixed-class proteins can induce LLPS through the combination of the interaction among LCRs, structural domains and nucleic acids.

NeProc predicts binding segments in intrinsically disordered regions without learning binding region sequences.

Intrinsically disordered proteins are those proteins with intrinsically disordered regions. One of the unique characteristics of intrinsically disordered proteins is the existence of functional segments in intrinsically dis-ordered regions. These segments are involved in binding to partner molecules, such as protein and DNA, and play important roles in signaling pathways and/or transcriptional regulation. Although there are databases that gather information on such disordered binding regions, data remain limited. Therefore, it is desirable to develop programs to predict the disordered binding regions without using data for the binding regions. We developed a program, NeProc, to predict the disordered binding regions, which can be regarded as intrinsically disordered regions with a structural propensity. We only used data for the structural domains and intrinsically disordered regions to detect such regions. NeProc accepts a query amino acid sequence converted into a position specific score matrix, and uses two neural networks that employ different window sizes, a neural network of short windows, and a neural network of long windows. The performance of NeProc was comparable to that of existing programs of the disordered binding region prediction. This result presents the possibility to overcome the shortage of the disordered binding region data in the development of the prediction programs for these binding regions. NeProc is available at http://flab.neproc.org/neproc/index.html.

Functional Segments on Intrinsically Disordered Regions in Disease-Related Proteins.

One of the unique characteristics of intrinsically disordered proteins (IPDs) is the existence of functional segments in intrinsically disordered regions (IDRs). A typical function of these segments is binding to partner molecules, such as proteins and DNAs. These segments play important roles in signaling pathways and transcriptional regulation. We conducted bioinformatics analysis to search these functional segments based on IDR predictions and database annotations. We found more than a thousand potential functional IDR segments in disease-related proteins. Large fractions of proteins related to cancers, congenital disorders, digestive system diseases, and reproductive system diseases have these functional IDRs. Some proteins in nervous system diseases have long functional segments in IDRs. The detailed analysis of some of these regions showed that the functional segments are located on experimentally verified IDRs. The proteins with functional IDR segments generally tend to come and go between the cytoplasm and the nucleus. Proteins involved in multiple diseases tend to have more protein-protein interactors, suggesting that hub proteins in the protein-protein interaction networks can have multiple impacts on human diseases.

Both Intrinsically Disordered Regions and Structural Domains Evolve Rapidly in Immune-Related Mammalian Proteins.

Eukaryotic proteins consist of structural domains (SDs) and intrinsically disordered regions (IDRs), i.e., regions that by themselves do not assume unique three-dimensional structures. IDRs are generally subject to less constraint and evolve more rapidly than SDs. Proteins with a lower number of protein-to-protein interactions (PPIs) are also less constrained and tend to evolve fast. Extracellular proteins of mammals, especially immune-related extracellular proteins, on average have relatively high evolution rates. This article aims to examine if a high evolution rate in IDRs or that in SDs accounts for the rapid evolution of extracellular proteins. To this end, we classified eukaryotic proteins based on their cellular localizations and analyzed them. Moreover, we divided proteins into SDs and IDRs and calculated the respective evolution rate. Fractional IDR content is positively correlated with evolution rate. For their fractional IDR content, immune-related extracellular proteins show an aberrantly high evolution rate. IDRs evolve more rapidly than SDs in most subcellular localizations. In extracellular proteins, however, the difference is diminished. For immune-related proteins in mammals in particular, the evolution rates in SDs come close to those in IDRs. Thus high evolution rates in both IDRs and SDs account for the rapid evolution of immune-related proteins.