graph neural network

Spanning-Tree Thermostatistics of Protein Allostery: An Exact Kirchhoff Framework with Application to Oncogenic KRAS

Kirchhoff matrix-tree evaluation of residue-contact spanning-tree ensembles yields exact allosteric thermodynamics; KRAS G12D conserves global and channel heat capacity but reroutes paths via Q61 and F156.

Calibration-Aware and Interpretable Graph Learning for Multi-Cohort Diffusion Connectome Brain-Age Modeling

This paper develops a calibration-aware graph learning framework for brain-age modeling across multiple cohorts, finding that recalibration improves performance but cohort-specific calibration is needed for external application.

InterPET: A Curated Benchmark of Sequence Embeddings and Graph Architectures with Interpretability and Biological Validation for PETase Activity Prediction

The paper introduces InterPET, a benchmark for predicting PETase activity using sequence embeddings and graph architectures, showing that model performance and biological validity are interdependent.

Real-time repository-scale spectral search and global molecular networking with HNSW-MS

HNSW-MS enables fast and accurate spectral search in large metabolomics repositories, allowing rapid integration of new spectra into global molecular networks.

A Biologically Informed Heterogeneous Graph Neural Network for Multi-Task Prediction of ncRNA-Metastasis-Cancer Interactions

The paper introduces a heterogeneous graph neural network to predict ncRNA interactions with cancer types and metastatic events, achieving high AUROC and F1 scores.

PandaDock: An Open-Source Molecular Docking Platform with Flexible-Ligand Search and Equivariant Neural Scoring

PandaDock is an open-source molecular docking platform that improves pose prediction and affinity scoring using a GNN and flexible-ligand search, achieving strong performance on benchmark datasets.

A Generative Virtual Tissue Model Enables Computational Design of Therapeutic Perturbation Strategies

This paper introduces CIFM, a generative virtual tissue model that simulates transcriptional dynamics under therapeutic conditions to design immunotherapy strategies for cancer and autoimmune diseases.

Learning Discrete Cell and Niche Codes from Spatial Transcriptomics Using Dual Residual Vector Quantization

SQUINT learns discrete cell and niche codes from spatial transcriptomics data using residual vector quantization, outperforming baselines in identification and cross-section integration.

Structure-Based TCR-pMHC Binding Prediction and Generalization to Unseen Peptides

This work evaluates factors affecting generalization of GNN-based TCR-pMHC binding predictors to unseen peptides and proposes architectural improvements.

Moirai: single-cell trajectory inference grounded in gene-level expression dynamics

Moirai infers cell trajectories by focusing on gene-level expression dynamics, improving detection of branching points in developmental processes.

Structure-aware Graph Learning Predicts RNA Editability Across Tissues and Species

AdarEdit predicts RNA editability using a structure-aware graph model trained on Alu duplexes across tissues and species, showing strong performance and evolutionary conservation.

Mapping disease traits onto spatiotemporal domain landscapes through scalable multi-sample integration of spatial transcriptomics

STUltra integrates spatial transcriptomics data to map disease traits onto spatiotemporal tissue landscapes, outperforming existing methods in scalability and speed.

A pangenome-graph approach for mapping and imputing barley sequences

Pan20, a Practical Haplotype Graph of 20 barley genomes, assigns sequences MorexV3 coordinates and imputes genotyping-by-sequencing (GBS) reads at 0.06% base-call error versus 0.32% for variant calling.

ClinOracle: Hierarchical AI Prediction of Target Binding and Patient-Derived Functional Activity Across Diverse Therapeutic Targets

ClinOracle uses a hierarchical graph neural network to predict target binding and patient-derived functional activity, improving compound prioritization through a Priority Score combining translational probability and developability.

DNA2Graph enables automated identification of non-linear DNA molecules in electron microscopy

DNA2Graph automates the identification of non-linear DNA molecules in electron microscopy images, significantly reducing the need for manual review and improving structure detection.

CellAwareGNN: Single-Cell Enhanced Knowledge Graph Foundation Model for Drug Indication Prediction

Adding OneK1K immune cell-type expression edges to an updated PrimeKG modestly improved drug-indication link prediction, AUPRC 0.816 to 0.826 overall and 0.847 to 0.864 for 40 autoimmune diseases.

Multi-modal Graph Integration for Biologically Interpretable Domain Identification Using Spatial Transcriptomics

This study introduces BIGraph-ST, a method that integrates pathway activity and histological features to improve spatial domain identification in spatial transcriptomics, showing better performance in cancer tissues.

A Physics-Inspired Classical Digital Twin of the Cell: A Composite Multi-Clock Port-Hamiltonian Neural Network Learned from Multi-Omic Circadian Data

The paper develops a physics-inspired digital twin of the cell using a multi-clock port-Hamiltonian neural network trained on multi-omic circadian data, showing aggregate phase lag agreement but no per-gene correlation.

MEGA-ODE: Learning Biologically Structured and Navigable Continuous Perturbation Dynamics from Sparse Omics

MEGA-ODE reconstructs dynamic omics landscapes from sparse data, predicts unmeasured states, and prioritizes perturbations toward biological endpoints.

MolX: A Geometric Foundation Model for Protein-Ligand Modelling

MolX is a geometric foundation model that jointly learns representations of protein pockets and ligands from 3D structural data, achieving state-of-the-art performance in predicting molecular interactions.

Graph Machine Learning for Physiological Role Prediction in Protein Contact Networks: A Large-Scale Comparative Study on the Human Proteome

This paper compares graph machine learning methods for predicting protein function from contact networks, finding GNNs outperform kernel methods in EC class prediction.

TriGraphQA: a triple graph learning framework for model quality assessment of protein complexes

TriGraphQA improves model quality assessment of protein complexes by using a triple graph framework that distinguishes intra-chain and inter-chain features, outperforming existing methods on benchmark datasets.

Subcellularly Resolved 3D Translatome in Mouse Oocytes and Early Embryos

The paper profiles the subcellularly resolved 3D translatome in mouse oocytes and early embryos, revealing spatially controlled translation patterns and embryo asymmetry.

MassGAT: a graph-based collective learning approach for untargeted detection and annotation of LC-MS data

MassGAT improves LC-HRMS data analysis by combining peak detection and annotation using a graph-based approach, outperforming existing methods on real datasets.

A brain dynamic model based on graph neural network reflect the inter-region interaction of cortical areas

This paper develops a graph neural network model to predict brain activity and quantify inter-regional interactions, showing it can distinguish Alzheimer's patients from controls.

Predicting Enzyme pH Optima from Structure Using Equivariant Graph Neural Networks

This paper introduces pHoptNN, a graph neural network predicting enzyme pH opt from structure, achieving lower error than sequence-based methods.

Quantitative assessment of cell fate commitment in single-cell transcriptomics using scCS

The paper introduces scCS, a framework for quantifying cell fate commitment parameters in single-cell transcriptomics, using methods like DFFP and providing diagnostics and visualizations.

Hypergraph geometry localizes mutation-associated dependency vulnerabilities in cancer.

The paper identifies mutation-associated therapeutic vulnerabilities in cancer using hypergraph geometry and CRISPR screens, finding specific dependencies with drug sensitivity.

PlasChain: an algorithm for improving long plasmid reconstruction from metagenome assemblies

PlasChain improves long plasmid reconstruction from metagenome assemblies by addressing fragmentation and false positives, outperforming existing methods in recovering long plasmids.

LNGCN: a distance-aware continuous-time graph framework for PPI prioritization in experimental screening

LNGCN improves PPI prioritization by using a distance-aware continuous-time graph framework, achieving high precision on imbalanced datasets and preserving residue-level representation diversity.