Skip to main content
Search

# Graph Transformer

4 related articles

Subscribe to updates
  1. NMR spectra now yield molecular skeletons directly: inverse-IMPRESSION solves 11 of 19 experimental molecules

    inverse-IMPRESSION maps NMR spectra directly onto bond probabilities between atoms, assembling candidate structures through one-shot prediction, stepwise bond filling and noise-augmented multishot sampling. Where comparable graph methods drop sharply on experimental spectra, the platform reaches 77.8% Top-1 on simulated data using only hydrogen and carbon NMR information and correctly identifies 11 of 19 experimental molecules, including a polycyclic steroid of 480.7 Da with an nSPS of 48.1.

  2. arXiv

    ZAGNet aggregates lung ultrasound on an anatomical zone-adjacency graph, reaching patient-level AUC 0.803 for consolidation and 0.893 for pleural effusion, above max and mean pooling

    The work presents ZAGNet, a zone-aware graph neural network that represents temporally tracked lung ultrasound pathology findings as graph nodes connected by anatomical zone adjacency, uses a graph transformer to propagate context across neighboring lung regions, and aggregates graph-level features through a virtual global node to predict patient-level consolidation and pleural effusion under patient-level supervision only; on a multicenter dataset of 714 subjects and 20,256 LUS video loops with exams spanning 4 to 16 zones across anterior, posterior, and lateral thoracic regions, AUC was 0.803 for consolidation (max pooling 0.677, mean pooling 0.674) and 0.893 for pleural effusion (max pooling 0.804, mean pooling 0.815), with missing zones accommodated by computing on a graph structure.

  3. bioRxiv

    A Graph-based QSAR Modeling Pipeline for Predicting In vitro PubChem Assays and In vivo Human Hepatotoxicity: Mechanistic Analysis of Caspase-3/7 Activation

    This study developed a graph-based QSAR modeling pipeline integrating assay data preprocessing, fingerprint and molecular graph feature representations, and benchmarking of classical machine learning, graph neural networks, graph transformers, and their consensus ensembles, applied to predict Caspase-3/7 activation, mitochondrial membrane potential disruption, and FDA drug-induced liver injury, where Graphormer achieved the highest F1 of 0.79 and the full consensus model achieved the highest AUC of 0.69 on DILI prediction, surpassing the previous best model with AUC 0.63 and F1 0.65, and identified structural motifs associated with dual activation and cell-line-specific responses through fragment enrichment analysis.

  4. SSE-DDI: Selective Substructure Encoding with Bond-Centered Molecular Representations for Drug-Drug Interaction Prediction

    This work proposes SSE-DDI, a framework that performs selective substructure encoding in molecular line graphs with chemical bonds as the fundamental representation units, complemented by an edge-fusion graph transformer and refined SMILES-derived Morgan-fingerprint similarity profiles; on DrugBank and Twosides under transductive and inductive settings it outperforms representative baselines across multiple metrics, with ablation and visualization analyses supporting the effectiveness of selective encoding and highlighting DDI-relevant molecular substructures.

You’re all caught up