Biology subjects

Kurgan, L.

Publications and source records attributed to Kurgan, L..

6 recordsLinked to original sources

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↗

bioRxiv2026

Relief of allosteric inhibition, redox imbalance, and transport limitations enables high-yield L-malate production in Escherichia coli

Malic acid is a C4 dicarboxylic acid traditionally produced from petroleum and widely used in the food industry. As a sustainable alternative, it can also be produced as a value-added platform chemical from biomass. Previously, the Escherichia coli strain XZ658 was engineered to produce L-malate via the carbon-fixation reductive branch of the TCA cycle. In this study, we further improved this system by relieving allosteric regulation of citrate synthase, addressing redox imbalance, and enhancing malate export. These modifications approximately doubled the L-malate titer in the final strain MO128 compared to XZ658 under simple batch fermentation conditions. The process achieved a high mass yield of 1.2 g malate g-{superscript 1} glucose, highlighting the carbon-fixation capacity of the reductive TCA pathway for fermentative malate production.

bioengineering↗

bioRxiv2026

Structure-aware geometric graph learning for modeling protease-substrate specificity at scale

Protease-substrate specificity is central to cellular regulation and disease pathogenesis, and accurately modeling its structural determinants remains challenging. Substrate recognition is governed by spatial constraints and higher-order relationships that extend beyond local sequence motifs. Most computational approaches rely predominantly on motif-centric or sequence-based representations, limiting their ability to capture the geometric and relational structure underlying enzymatic specificity. Here, we introduce OmniCleave, a structure-aware geometric graph learning framework for modeling protease-substrate specificity at scale. OmniCleave is trained on 57,278 structure-informed protease-substrate pairs derived from 9,651 substrates spanning over 100 proteases across six distinct families. The framework integrates multi-scale structural graphs with higher-order protease relational topology, explicitly encoding spatial context and inter-protease dependencies within a unified geometric representation. This formulation moves beyond local pattern recognition and enables transferable modelling across six protease families. Across large-scale benchmarks, the framework consistently outperforms existing approaches and reveals interpretable geometric determinants underlying substrate recognition. Experimental validation confirms three novel caspase-3 substrates and 21 cleavage sites predicted by OmniCleave, supporting the biological relevance of the learned representations. Together, OmniCleave provides a scalable geometric framework for modeling protease-substrate specificity, with practical utility for systematic analysis of protease biology.

bioinformatics↗

bioRxiv2025

KcatNet: advancing genome-wide enzyme turnover number prediction through structural enzymatic characterization

Enzyme turnover numbers (Kcat) are fundamental kinetic constants that quantify enzymatic efficiency. Systematic studies of Kcat are essential for characterizing the mechanisms underlying proteomic composition and cellular metabolism. However, experimental measurements of Kcat remain limited and prone to noise. To address this, we present KcatNet, a geometric deep learning model designed for high-throughput prediction of Kcat in metabolic enzymes across all organisms, leveraging paired enzyme sequence and substrate representations. KcatNet consistently outperforms existing predictors, particularly for enzymes with high catalytic efficiency, and demonstrates strong generalization to enzymes that are dissimilar to those in the training set. Furthermore, KcatNet uncovers structural mechanisms and interaction patterns within enzyme-substrate complexes, providing insights into architectural principles that are inaccessible with existing methods by harnessing the representational power of large-scale protein language models. We applied KcatNet to genome-scale Kcat prediction across diverse yeast species, improving proteome allocation predictions by integrating its outputs into metabolic models. Experimental validation further confirmed the models ability to identify enzyme mutants with enhanced activity. By bridging the gap between sequence, structure, and function, KcatNet establishes a robust foundation for advancing understanding of molecular-level mechanisms and accelerating enzyme engineering efforts.

bioinformatics↗

bioRxiv2025

PROBind: A Web Server for Prediction, Analysis and Visualization of Protein-Protein and Protein-Nucleic Acid Binding Residues

Protein-protein and protein-nucleic acids interactions are fundamental to numerous cellular functions, yet only a small fraction have been experimentally characterized. Although modern computational methods have been developed for predicting interacting residues in proteins, they are challenging to use due to individual installation and execution requirements, lack of a standardized input or output format, and absence of support for result analysis. Moreover, methods trained using structures of complexes or intrinsically disordered regions, may not perform well on other types. To overcome these challenges, we develop PROBind, a web server for predicting, analyzing, and interactively visualizing protein, DNA and RNA binding residues from both protein sequences and structures. PROBind integrates 12 predictors trained on structural or disordered proteins, and supports the upload of results from external predictors. By normalizing and averaging predictions from multiple predictors targeting the same ligand type, PROBind generates meta-predictions that balance discrepancies among different methods. Furthermore, it provides interactive graphical tools for result analysis and contextualization. Overall, PROBind accommodates diverse ligand types and supports predictions and analysis based on both structure and sequence data, overcoming the limitations of existing tools. PROBind is freely accessible at https://www.csuligroup.com/PROBind.

bioinformatics↗

bioRxiv2024

SCREEN: a graph-based contrastive learning tool to infer catalytic residues and assess mutation tolerance in enzymes

The accurate identification of catalytic residues contributes to our understanding of enzyme functions in biological processes and pathways. The increasing number of protein sequences necessitates computational tools for the automated prediction of catalytic residues in enzymes. Here, we introduce SCREEN, a graph neural network for the high-throughput prediction of catalytic residues via the integration of enzyme functional and structural information. SCREEN constructs residue representations based on spatial arrangements and incorporates enzyme function priors into such representations through contrastive learning. We demonstrate that SCREEN (i) consistently outperforms currently-available predictors; (ii) provides accurate results when applied to inferred enzyme structures; and (iii) generalizes well to enzymes dissimilar from those in the training set. We also show that the putative catalytic residues predicted by SCREEN mimic key structural and biophysical characteristics of native catalytic residues. Moreover, using experimental data sets, we show that SCREENs predictions can be used to distinguish residues with a high mutation tolerance from those likely to cause functional loss when mutated, indicating that this tool might be used to infer disease-associated mutations.

bioinformatics↗