1012 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.
This paper introduces Atlas, a large-scale microbiome pretraining dataset, and the Waypoint family of foundation models, demonstrating through the Compass benchmark that self-supervised pretraining significantly outperforms classical methods and existing models in diverse microbiome prediction tasks.
This paper identifies and characterizes systematic contextual biases in the SegmentNT nucleotide transformer model—specifically regarding input sequence length, nucleotide position, and a 24-nucleotide periodic oscillation linked to tokenization—and proposes standardization methods to improve prediction consistency and guide the use of similar genomic models.
MilliMap is an interactive framework that unifies statistical computation and spatial exploration to enable closed-loop, iterative analysis of spatial omics data, allowing researchers to refine parameters and validate findings within a single environment for applications ranging from neuroanatomy to tumor microenvironments.
This paper presents the SenNet Data Portal, a scalable hybrid cloud platform that provides open access to harmonized multimodal, multi-tissue datasets and analytical tools to systematically characterize cellular senescence across human and mouse models for aging research.
This benchmark study challenges the assumption that larger AI models universally outperform smaller ones in drug discovery, demonstrating that compact, specialized models often achieve superior or comparable predictive accuracy across diverse molecular property and activity tasks compared to large foundation models.
AnnotateMissense is a scalable framework that integrates diverse genomic and protein-language-model features to benchmark and generate high-performance pathogenicity predictions for over 90 million missense variants, achieving superior accuracy with an XGBoost model trained on 132,714 ClinVar-labeled variants.
The authors developed the Structural Novelty Index (SNI), an AI-driven framework that successfully identified and experimentally validated an unexpected undecameric assembly of NRC immune receptors, demonstrating a scalable method for discovering atypical protein complexes beyond canonical architectures.
This paper proposes an unsupervised framework leveraging large language models and contrastive learning to characterize SARS-CoV-2 spike protein sequences, demonstrating improved clustering performance in predicting emerging variants compared to previous approaches.
The paper introduces cryoPANDA, a massive dataset of over 37 million annotated cryo-EM particles from 252 diverse experiments, designed to overcome current data limitations and accelerate the development of data-driven methods for structural biology.
This paper presents a fully data-driven subspace one-class classification framework that models healthy plasma proteomic profiles to robustly detect diverse diseases without requiring diseased training samples, thereby overcoming class imbalance challenges in high-dimensional clinical data.