Biology subjects

Malhis, N.

Publications and source records attributed to Malhis, N..

4 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↗

bioRxiv2025

Predicting molecular recognition features in protein sequences with MoRFchibi 2.0

Molecular Recognition Features (MoRFs) are segments within disordered protein regions (IDRs) that undergo a disorder-to-order transition upon binding to their partners. Identifying MoRFs remains a significant challenge. This paper introduces MoRFchibi 2.0, a specialized prediction tool designed to identify the locations of MoRFs within protein sequences. Our results show that MoRFchibi 2.0 outperforms all existing MoRF and general predictors of protein-binding sites within IDRs, including top-performing models from CAID rounds 1, 2, and 3. Remarkably, MoRFchibi 2.0 surpasses predictors that utilize AlphaFold data and state-of-the-art protein language models, achieving superior ROC and Precision-Recall curves and higher success rates. MoRFchibi 2.0 generates output scores using an ensemble of logistic regression convolutional neural network models, followed by a reverse Bayes Rule to adjust for priors in the training data. These scores reflect MoRF probabilities normalized for the priors in the training data, making them individually interpretable and compatible with other tools utilizing the same scoring framework. AvailabilityAn online server: https://mc2.msl.ubc.ca/index.xhtml and code: https://github.com/NawarMalhis/MC2.git.

bioinformatics↗

bioRxiv2024

Probabilistic Annotations of Protein Sequences for Intrinsically Disordered Features

This paper introduces a novel platform for IDR Probabilistic Annotation (IPA). The IPA platform now encompasses tools for predicting Linker regions and nucleic, protein, and all (protein or nucleic) IDR binding sites within protein amino acid sequences. Despite its simplicity and computational efficiency, results demonstrate that IPA performs competitively with leading tools in predicting protein and all IDR binding sites while considerably outperforming all tools in identifying Linker regions and nucleic binding sites. An important contribution of this work is the introduction of a new output paradigm for computational feature predictions. Traditional tools typically express predictions as scores, with higher values indicating greater probabilities. However, these scores lack true probabilistic meaning and interpretability, even derived from logistic regression models. This limitation arises primarily because training data priors differ from broader populations unknown priors. This paper proposes applying a reverse Bayes Rule to logistic regression outputs, effectively normalizing for the priors in the training data. This adjustment produces scores representing actual probabilities, assuming 50% priors in the general population. Such scores are interpretable in isolation and enable comparability and integration across different tools, marking a significant step toward standardization in feature prediction methodologies. Availabilityorca.msl.ubc.ca/nmshare/ipa.tar.gz

bioinformatics↗

bioRxiv2024

Pre-processing annotated homologous regions in protein sequences concerning machine-learning applications

Accurate preprocessing of annotated protein sequences with regard to homologies is essential for maintaining the integrity of machine-learning applications. This study presents two new tools--HAM (Homology-based Annotation Masking) and HAC (Homology Annotation Conflict)-- designed to address these challenges. HAM detects and masks homologous regions between datasets to prevent leakage, while HAC identifies and resolves annotation inconsistencies within datasets. Applying these tools to three benchmark datasets revealed substantial overlooked homology and annotation conflicts, even in datasets that had been previously clustered by sequence identity. These findings underscore the importance of homology-aware preprocessing to ensure the integrity of model training and evaluation. By integrating HAM and HAC into machine learning workflows, researchers can improve the consistency and trustworthiness of protein sequence-based predictions. Availabilitygithub.com/NawarMalhis/HAM.git

bioinformatics↗