Biology subjects

Erdos, G.

Publications and source records attributed to Erdos, G..

4 recordsLinked to original sources

bioRxiv2026

DEEP: Extracting Latent Energetic Landscapes from Protein Structure Predictors

Modern protein structure predictors learn complex sequence-structure relationships, yet the extent to which their internal representations capture the energetic principles governing protein stability and conformational preferences remains unclear. Here, we introduce DEEP (Distogram Encoded Energetic Potential), a general framework that extracts effective statistical potentials directly from predicted residue-pair distance distributions, using the models own terminal distance state as an internal reference without training on experimental stability data. Applied to Boltz-2 (DEEP-Boltz) achieves a mean per-assay Spearman correlation of 0.720 across 64 ProteinGym stability assays, numerically outperforming all 97 published zero-shot methods in the matched comparison on level with the best performing predictors trained directly on experimental stability measurements. Beyond mutation effects, DEEP-Boltz accurately distinguishes native-like from non-native conformations across more than 180,000 Rosetta decoys. Applying DEEP to AlphaFold2 reveals nearly identical conformational discrimination but substantially weaker mutation-effect prediction, demonstrating that recognizing favorable conformations and capturing their energetic sensitivity to sequence changes are distinct properties of learned structural representations. Together, these findings establish that protein structure predictors encode accessible energetic information beyond their geometric predictions and that this information can be extracted. We release the framework and DEEP-Boltz as a freely available tool for extracting these latent energetic representations and for zero-shot prediction of mutation-induced protein stability changes.

bioinformatics↗

bioRxiv2026

The Hidden Disorder Divide: Reconciling Benchmark Inconsistencies in Intrinsically Disordered Protein Binding Site Prediction

Computational predictors of protein-binding sites within intrinsically disordered regions (IDRs) show highly inconsistent performance across high-quality benchmark datasets. To understand the origins of these discrepancies, we systematically compared predictors across three independent test sets: two CAID datasets updated with the latest DisProt annotations and a composite dataset (DBs) assembled from DIBS, FuzDB, IDEAL, and MFIB. Predictors trained predominantly on DisProt data achieved substantially higher AUCs on the CAID sets but performed poorly on the DBs. In contrast, predictors trained on older, low-quality PDB-based datasets showed balanced performance across all sets, with a slight preference for DBs. Predictors with mixed training exposure displayed intermediate behavior. Through controlled experiments using identical CNN architectures and feature analysis, we demonstrate that the dominant factor driving these performance differences is the intrinsic disorder propensity of the binding sites themselves. Binding residues in DisProt-based datasets exhibit markedly higher average disorder propensity scores than those in PDB-derived datasets. This previously unrecognized selection bias -- literature studies preferentially characterizing more disordered binding sites, while PDB-derived annotations capture less disordered ones -- effectively splits IDR-protein binding sites into two distinct categories. Predictors optimized on one category therefore generalize poorly to the other. Binding-site length and sequence conservation play only minor or negligible roles in explaining the observed inconsistencies. These findings highlight a critical limitation in current benchmarking practices and training strategies for IDR-binding site prediction, underscoring the need for more balanced and disorder-aware reference datasets. Finally, the diagnostic techniques introduced here could prove valuable beyond the specific application examined in this study. The data and code used to generate all figures and tables are available at: https://github.com/NawarMalhis/HDD

bioinformatics↗

bioRxiv2025

Pathogenic Missense Mutations in Intrinsically Disordered Regions Reveal Functional and Clinical Signals

Intrinsically disordered regions (IDRs) play key roles in cellular signaling and regulation, yet their contribution to human disease remains poorly understood. Here we analyzed nearly one million ClinVar missense variants, focusing on those located within IDRs defined by curated and predicted annotations. We found that pathogenic variants were significantly enriched in short linear motifs (SLiMs) and disordered binding regions, consistent with their central functional importance. To extend these insights beyond existing annotations, we applied AlphaMissense, a deep-learning pathogenicity predictor, and uncovered localized "island-like" patterns of elevated pathogenicity within IDRs. Leveraging these signals, we developed a classifier to prioritize predicted ELM motifs (PEMs), revealing thousands of candidate functional sites linked to major disease classes, including neurological, cardiovascular, and cancer-associated genes. Case studies in POLK, FOXP2, and LMOD3 illustrate how this framework connects genetic variation to molecular mechanisms, providing a scalable route to interpret variants of uncertain significance and advancing our understanding of pathogenicity in the disordered proteome. SummaryThis study reveals how deep-learning pathogenicity predictions can uncover functional motifs within intrinsically disordered regions, providing a new framework for interpreting genetic variation in the disordered proteome.

bioinformatics↗

bioRxiv2018

Large-scale analysis of redox-sensitive conditionally disordered protein regions reveal their widespread nature and key roles in high-level eukaryotic processes

Recently developed quantitative redox proteomic studies enable the direct identification of redox-sensing cysteine residues that regulate the functional behavior of target proteins in response to changing levels of reactive oxygen species (ROS). At the molecular level, redox regulation can directly modify the active sites of enzymes, although a growing number of examples indicate the importance of an additional underlying mechanism that involves conditionally disordered proteins. These proteins alter their functional behavior by undergoing a disorder-to-order transition in response to changing redox conditions. However, the extent to which this mechanism is used in various proteomes is currently unknown. Here, we use a recently developed sequence-based prediction tool incorporated into the IUPred2A web server to estimate redox-sensitive conditionally disordered regions on a large scale. We show that redox-sensitive conditional disorder is fairly widespread in various proteomes and that its presence strongly correlates with the expansion of specific domains in multicellular organisms that largely rely on extra stability provided by disulfide bonds or zinc ion binding. The analyses of yeast redox proteomes and human disease data further underlie the significance of this phenomenon in the regulation of a wide range of biological processes, as well as its biomedical importance.

bioinformatics↗