1197 papers
Bioinformatics sits at the exciting intersection where biology meets data science, using powerful computer tools to decode the vast complexity of living systems. From mapping the human genome to tracking how viruses evolve, this field transforms raw biological information into actionable insights that drive modern medicine and research forward without requiring a supercomputer to understand the basics.
On Gist.Science, we ensure you never miss a breakthrough by processing every new preprint in this category directly from bioRxiv. Our team provides both plain-language explanations and detailed technical summaries for each paper, making cutting-edge discoveries accessible to everyone regardless of their background.
Below are the latest bioinformatics papers added from bioRxiv, ready for you to explore with clarity and depth.
RegimeFormer is a large-scale protein model that, through its associated RegimeAtlas database of over 200 million sequences, establishes global perturbation regimes to predict residue-level mutation effects and improve downstream biological modeling, particularly for unseen proteins and families.
This paper introduces the Dynamic Hierarchical Interleaved Bloom Filter, a scalable and updatable indexing structure that extends the state-of-the-art HIBF with partial rebuilding to enable efficient large-scale sequence search, demonstrating the ability to index over 100 TB of RNA-Seq data and insert new samples 24 to 65 times faster than competing tools.
Vipsania is the first unsupervised deep learning tool that accurately predicts protein-coding gene structures across diverse eukaryotic genomes by learning directly from unannotated sequences, thereby overcoming the limitations of supervised methods that require high-quality training data and struggle with distant or basal clades.
PhageTransformer is a deep learning model that overcomes the limitations of existing tools by providing scalable, accurate, and rapid host predictions for bacteriophages, as demonstrated by its superior performance on thousands of independent phage-host pairs.
The paper introduces morphOTU, an AI-driven framework that derives operational biodiversity units directly from specimen images using self-supervised learning and hierarchical clustering, enabling accurate quantification of species diversity and discovery of morphological patterns even in the absence of formal taxonomic labels or extensive training data.
The paper introduces DeepTMHMM2, a novel predictor that simultaneously achieves accurate transmembrane topology prediction—including previously unmodeled re-entrant regions and interfacial helices—and subcellular localization across 17 biological membranes, revealing the widespread presence of non-crossing segments in the proteome.
This study presents an updated XMAn database that integrates extensive COSMIC mutation data to create specialized peptide repositories, enabling the mass spectrometry-based identification of cancer-associated mutated peptides that are typically missed by standard canonical protein databases.
This paper presents a hybrid framework that combines a Variational Autoencoder to generate synthetic longitudinal trajectories from cross-sectional kidney cancer data with a symbolic rule-induction system to produce interpretable, probabilistic forecasts of disease progression that match deep learning accuracy while offering transparent, human-readable insights into the underlying molecular mechanisms.
HIDE-Deconv is an open-source hierarchical deconvolution framework that jointly optimizes cellular composition estimates across multiple resolution levels to reveal biologically relevant cellular remodeling in diseases like lung adenocarcinoma and sepsis, outperforming existing single-resolution methods.
The paper introduces evoPLM-Tree, a tree-aware conditional language model that integrates phylogenetic context to accurately predict viral mutational patterns and lineage-specific evolutionary trajectories, demonstrating strong alignment with experimental functional constraints.