large language model

Towards an analytical framework for investigating the neurophysiological influence of social context in speech perception

The paper introduces a framework combining a fine-tuned LLM and TRFs to isolate social context's influence on speech perception, showing social context improves EEG prediction.

FeatureMAP enables network-guided interpretation of untargeted metabolomics

FeatureMAP improves interpretation of untargeted metabolomics data by integrating annotation, network refinement, and functional analysis, revealing metabolic patterns in Alzheimer's disease.

SPaRTA: Spatial Reasoning in Text-to-SQL for Omics Analysis

SPaRTA enables non-experts to query spatial-omics data via natural language, achieving 63-77% accuracy with fine-tuning on cancer tissue datasets.

DEEPNET-LLM: Highly accurate structure prediction of unknown metabolites from mass spectra via their molecular neighborhoods

DEEPNET-LLM predicts unknown metabolite structures from mass spectra by leveraging molecular neighborhoods, achieving high accuracy in de novo structure prediction.

Small molecule structure elucidation from NMR spectra with generative and reasoning models

This paper introduces an AI system for automated small molecule structure elucidation from NMR spectra, combining generative models with reasoning agents to outperform existing tools.

Agentic Discovery Transfer Enables Cross-Cancer Discovery at Scale

SMART, a multi-agent framework, transfers findings from single-cell RNA-sequencing cancer studies to other cancer types, identifying cross-cancer prognostic signatures and generating new hypotheses.

An evidence-governed agent harness for spatial transcriptomics prioritizes ICAM-1 as a target in lung fibrosis

DeepMAST, an AI system for spatial transcriptomics, identifies ICAM-1 as a therapeutic target in lung fibrosis through evidence-based analysis and validation.

What comes after de novo? Automated lead optimization of proteins with CRADLE-1

CRADLE-1 automates protein lead optimization, improving speed and efficiency compared to rational design methods.

From Closed-Set Annotation to In-Context Discovery: Biologically Grounded LLMs for Cellular Profiling

CellScribe is a generative framework that unifies cell identity and state by using a language model to generate cell type annotations and meta-program profiles, outperforming baselines and identifying new cell types through in-context learning.

Towards Robust Automated Mouse Scratching Behavior Detection for Itch Research

This paper develops Scratch-AID+, an automated mouse scratching detection framework, showing that fine-tuned video foundation models outperform classic models in generalizability and accuracy.

CLARA: Compound Library Active Learning with Reasoning Agents

CLARA uses a language model to make strategy decisions in compound library active learning, outperforming random sampling across multiple protein targets.

Inherent limitations of zero-shot mutation prediction in protein engineering

This study evaluates zero-shot mutation prediction models in protein engineering, finding they struggle to identify optimal mutations and predict new functions due to limitations in capturing evolutionary constraints.

Large language models unlock the ecology of species interactions

This paper uses large language models to extract species interaction data from citizen science comments, expanding ecological data coverage and enabling new ecological research.

FragDockRL: A Reinforcement Learning Method for Fragment-Based Ligand Design via Building Block Assembly and Tethered Docking

FragDockRL uses reinforcement learning to design ligands by assembling building blocks and tethered docking, outperforming random search on three protein targets.

Agentic campaign control for high-throughput de novo binder design

This paper introduces T-REX, an agentic framework for protein binder design that adaptively allocates compute resources across methods, achieving higher throughput of structurally distinct hits than baselines.

Drought-Spec-Net: Early Tomato Drought Detection and Potential Yield-Impact Assessment Using Vis NIR Data

This paper develops Drought-Spec-Net, a model for early tomato drought detection using Vis NIR data, achieving high accuracy and providing yield impact assessments.

RNASeek: A Cross-Phyla Generative Foundation Model for Multipurpose RNA Modeling and Reinforcement Learning-Based Design

RNASeek is a generative foundation model for RNA sequence design and functional prediction, using reinforcement learning to optimize RNA properties like stability and cleavage activity.

BOTANIC-1: a series of long-context plant genomic foundation models in the agentic era

Bidirectional Mamba2 plant genomic language models (318M to 3.2B parameters) trained on 320 land-plant genomes top a 22-task frozen-model aggregate; a small variant reaches 131,072 bp context.

BoneGraph: A Domain-Specialised, Self-Correcting Reasoning System for Bone Science Retrieval, Grounded Inference, and Image Mechanics

BoneGraph combines a SPECTER2-indexed bone corpus, knowledge graph, deterministic physics checks, and user-taught rules in a local five-tab system; oracle passages raise accuracy 42% to 78%.

SemVac: A Semantic Vaccinology Paradigm Powered by LLMs for Antigen Discovery

SemVac uses LLMs to predict vaccine antigens by analyzing literature-derived protein descriptions, outperforming existing methods and identifying new candidates with experimental validation.

EukaUTR: a foundation model unifying functional modelling and design of eukaryotic 3' UTRs

EukaUTR is a foundation model for eukaryotic 3' UTRs that predicts function and generates new sequences with regulatory properties.

SCALE: Scalable Conditional Atlas-Level Endpoint transport for virtual cell perturbation prediction

A set-aware latent transport model predicts treated cell populations from unpaired controls, keeps perturbation-specific signal in CRISPR, drug and cytokine data, and guided a cytokine choice tested in three donors.

A low-dimensional, generalizable encoding manifold for auditory cortex

The paper develops ACNet, a foundation model of auditory cortex sound encoding, showing it can predict neural activity across animals and reveal generalizable coding principles.

De novo design of ligand binding proteins using large language models alone

This paper demonstrates that large language models can generate de novo protein sequences that bind specific ligands without copying existing sequences, with a 25% hit rate in experimental validation.

Cellquant: a vibecoder's guide to image analysis

One Python script, configured through LLM-assisted command-line calls, segments and quantifies 2D/3D multi-channel fluorescence images with provenance records; in a yeast temperature series it exposed a non-responding subpopulation.

Visual LLM-guided consensus spatial domain detection with L-STAR

L-STAR improves spatial domain detection by combining visual LLM guidance with consensus-based integration of multiple methods across diverse datasets.

ResLit: A Large-Scale Automated Literature Mining Database for Antimicrobial Resistance

ResLit couples BioMistral-7B screening with two-pass Qwen3-30B extraction to build an antimicrobial-resistance database of 3,120 genes and 13,593 mutations, evidence-tiered against CARD, ResFinder, and NCBI RGC.

Spatial second-order features predict glioma malignant transformation

This study identifies second-order spatial features, such as ligand-receptor interactions, as predictive of glioma malignant transformation using integrated spatial DNA and RNA profiling.

ACCREDIT: A Quality-Aware Agentic Engine for Cell-resolved Cross-modal Image Registration with Dynamic Iterative Tuning

ACCREDIT improves cross-modal image registration by detecting failures and automatically recovering alignments, outperforming existing methods in accuracy and robustness.

A cross-modal generative model for incomplete and degradedprostate MRI with multicentre clinical validation

This paper develops MSCNet, a cross-modal generative model for reconstructing and restoring prostate MRI sequences, showing improved image quality and diagnostic performance in multicentre validation.