Bioinformatics has evolved from command-line tools to AI-driven systems, with deep learning models like AlphaFold transforming structural biology and data interaction.
- 1 opens in JClub
- Open access
Oxford University Press · Genomics & Bioinformatics · ISSN 1467-5463, 1477-4054
Rebuilt
Bioinformatics has evolved from command-line tools to AI-driven systems, with deep learning models like AlphaFold transforming structural biology and data interaction.
Transformer-based language models have the potential to revolutionize bioinformatics by addressing complex challenges across genomics, transcriptomics, proteomics, and drug discovery.
Quantum bioinformatics shows promise in tackling complex biological data problems, with 10 key domains revealing emerging trends and methodological patterns.
Negative data augmentation with region-directed mutations improves TCR-epitope binding and interaction prediction, enhancing model discrimination and robustness.
Dual integrative genomic prediction framework improves cold stress tolerance prediction in wheat, achieving higher accuracy than traditional models across diverse datasets.
Newest first · last 60 days
bioETH-PRS enables confidential polygenic risk scoring entirely within encrypted data using smart contracts on an fhEVM-enabled blockchain, ensuring privacy and transparency.
DeepIR-Pred accurately identifies insulin receptor proteins with over 10-fold improved robustness and generalization using deep recurrent learning and biologically informed features.
Halo achieves over 20% higher accuracy than existing methods in whole-cell segmentation across diverse tissues using only nuclear images and RNA data.
Fundamental data quality issues were confirmed in enzyme datasets, with discrepancies found in several entries used for machine learning prediction of enzyme optimal pH.
Pathway-level minimum reaction fluxes outperform aggregate pathway activity in personalized metabolic modeling, achieving high diagnostic accuracy across diverse human diseases.
BioCon reveals that 35% of bioinformatics papers exhibit inconsistencies between descriptions and code implementations, highlighting significant gaps in research communication.
Non-Markovian Gillespie algorithms reveal that history-dependent reaction times can significantly alter biological population predictions, challenging traditional memoryless models.
Contrastive learning produces interpretable adverse event vectors that outperform existing methods in drug-event association prediction with an AUC of 0.88.
Over 80 bioinformatics tools, achieving over 0.95 Matthews correlation in identification, now support comprehensive bacteriophage research across multiple computational paradigms.
PMPIHGLL achieves over 0.9 AUC and AUPR in predicting metabolite-protein interactions across multiple datasets, outperforming existing models.
Machine learning enables identification of ncRNA signatures with high predictive accuracy for cancer diagnosis and prognosis, advancing early detection efforts.
FlexBIP, a modular framework for biomolecular interaction prediction, outperforms 25 state-of-the-art models across 15 datasets, including cold-start scenarios.
Amplification bias in spatial transcriptomics can distort gene expression data, with errors reaching up to significant levels that hinder accurate spatial analysis.
Machine learning models using consensus functions achieve high accuracy in predicting T-cell cross-immunity with small peptide datasets.
lagCI accurately infers temporal causal relationships from dense multi-omic time series, identifying over 157,000 interactions in human data.
Cell2space accurately reconstructs multi-scale tissue architecture from scRNA-seq and spatial transcriptomics data, achieving superior domain and neighborhood inference.
Masked-multi-layer perceptron improves robustness of genetic risk scores, achieving a Spearman correlation of 0.951 with noisy data compared to 0.669 for traditional methods.
Hidden causality analysis uncovers prevalent dynamic lncRNA regulatory patterns in autism spectrum disorder, involving 20 immune-related lncRNAs and eight susceptibility genes.
MMP2Mol enhances ligand-based drug design by increasing molecular novelty to over 79%, outperforming baseline models across multiple therapeutic targets with limited data.
DDTRN predicts bacterial transcriptional regulatory networks with high accuracy, achieving an average AUC of 0.869 and PR AUC of 0.868 across diverse species.
Long-read sequencing-based detection of germline and somatic structural variations achieves high accuracy with specific tools, with some methods reaching stable performance across diverse datasets.
Subtypist identifies novel cell subtypes in scRNA-seq data with over 30% improved accuracy compared to existing methods, without relying on external references.
ST-ConMa, a multimodal foundation model for spatial transcriptomics, outperforms existing methods by learning high-quality image-gene representations across diverse data.
Cell type-specific gene regulatory networks (GRNs) can be accurately reconstructed by integrating bulk and single-cell RNA-seq data, with CTN achieving high inference accuracy.
ToxiCompass achieves near-perfect accuracy in short-peptide toxicity screening, with an AUC of 0.9871 and a Hit@3 of 0.8955, enabling reliable assay prioritization.
Journals this month
Moving areas, week to 3 Oct 2026