The front page of AI-driven biology
Thursday, October 15, 2026Thu, Oct 15
Updated 11:02 PM PT
Recent original research from bioRxiv and arXiv. Preprints have not been peer reviewed.
CAMP-14-3-3 is a computational framework that predicts binding sites on 14-3-3 proteins, which regulate cellular signaling by docking to phosphorylated serine and threonine residues.
A team led by Tsinghua University introduced MicroGlot, a genomic language model pretrained on 378.3 billion nucleotides across 99,700 microbial species.
Researchers built an open resource of 2,986 metaphase fluorescence microscopy image sets with manual annotations to automate the measurement of extrachromosomal DNA, which are circular DNA fragments that drive tumor diversity and drug resistance.
A machine-learning model called PanPhage combines gene content from both bacteria and bacteriophages to predict whether a given virus can kill clinical strains of Escherichia coli.
Researchers trained a transformer neural network named ASOFormer on more than 170,000 antisense oligonucleotide target pairs to predict gene knockdown efficiency directly from RNA sequence, secondary structure, and synthetic chemical modifications.
Researchers have built Cerebra-Epistasis, a computational tool that combines single-sequence 3D protein structure prediction with an epistasis model to forecast how combinations of multiple mutations alter protein function.
Computational biologists paired elastic network models, which treat connected parts of a protein like tiny springs, with simplified Brownian dynamics simulations to map protein flexibility at a fraction of standard computing costs.
Researchers built ProFinder, a computational tool designed to identify reliable constitutive promoters, the DNA control switches that keep genes switched on, in unstudied bacterial species.
Researchers have developed Brain Landscape In Situ Crosslinking Mass Spectrometry, or BLIS-XL-MS, to profile proteome-wide protein conformational shifts and complex remodeling directly in intact brain tissue.
A computational workflow predicts how effectively bacterial beta-lactamase resistance enzymes degrade carbapenem antibiotics using only the enzyme's sequence as input.
Researchers have developed PepSpace, a computational platform that pairs coarse-grained molecular dynamics simulations with a genetic algorithm to design peptide binders from scratch.
Researchers fine-tuned large language models on a small clinical corpus to create Dementia Language Models that generate realistic narratives reflecting cognitive impairment.
Archival duodenal biopsies classified as having no pathologic abnormality show significant clinical and gene expression variations depending on whether a child has inflammatory disease elsewhere in the digestive tract.
A combined imaging workflow links Xenium spatial transcriptomics with iterative indirect immunofluorescence imaging on the same section of preserved human brain tissue.
Researchers mapped single-cell gene expression and paired T- and B-cell receptor sequences from the blood of 249 individuals with inflammatory bowel disease or healthy controls.
Researchers used structural modeling from AlphaFold3 to engineer dual-excitation ratiometric biosensors that track neuropeptides in live animals without interference from blood flow, tissue movement, or pH shifts.
Target-adaptive Rescue-Explore-eXploit, or T-REX, uses large language model agents to orchestrate multiple generative models and structure-evaluation tools during protein binder design campaigns.
Researchers introduced Minerva, an alignment-free framework that uses a genome language model to predict RNA base-pairing, protein contacts, and repeat sequences as two-dimensional interaction maps.
Researchers combined physical simulations with a generative machine-learning model, called latent thermodynamic flows, to map the shifting shapes and folding speeds of flexible RNA molecules.
Researchers have built a molecular foundation model that predicts cryptic binding pockets directly from unliganded, static protein coordinates without running structural simulations.
Researchers built the InterVir atlas using thermal proximity coaggregation to track how protein-protein interaction networks shift over time across five RNA viral pathogens.
DeepSCENIC transfers representations from sequence-to-function deep learning models, such as Enformer and Borzoi, to single-cell multiomic atlases to reconstruct causal gene regulatory networks.
Traffickome maps 754 transport proteins into 11 directed transitions across seven cellular compartments to infer where transmembrane cargo moves using standard proteomic profiles.
Computational biologists created cytokine prodrugs designed from scratch that stay inactive until cleaved by tumor-specific enzymes.
A graph-guided autoencoder called MeRN inferred metabolic pathway activity from single-cell RNA sequencing data to pinpoint why folate deficiency disrupts mouse neural tube closure.
Researchers built an automated design pipeline that translates specified analog computing behaviors into synthetic DNA circuits for mammalian cells.
A mechanistic mathematical model tracking both immune cell infiltration and circulating tumor DNA (ctDNA) sheds light on pseudoprogression, a phenomenon where immunotherapy causes tumors to temporarily swell before shrinking.
Peptides computationally engineered purely to change their secondary structure in response to environmental cues proved to act as potent antimicrobial agents in lab tests.
A language-model-guided computational tool identified specific guanine motifs that govern where RNA polymerase II pauses near gene promoters.
Researchers built a graph neural network that embeds evolutionary trees directly into its architecture, improving predictions of animal host origins for paramyxoviruses.