contrastive learning

HuiduRep: A Robust Self-Supervised Framework for Learning Neural Representations from Extracellular Recordings

HuiduRep is a self-supervised framework that learns robust representations from extracellular recordings, improving spike sorting performance across datasets and outperforming existing tools.

NeuroNarrator: A Generalist EEG-to-Text Foundation Model for Clinical Interpretation via Spectro-Spatial Grounding and Temporal State-Space Reasoning

NeuroNarrator translates EEG segments into clinical narratives using spectro-spatial grounding and temporal state-space reasoning, enabling interpretable clinical interpretation.

Rigidity-Aware Geometric Pretraining for Protein Design and Conformational Ensembles

This paper introduces RigidSSL, a geometric pretraining framework for protein design and conformational modeling, improving designability and capturing realistic dynamics.

Grounding olfactory perception in language: Benchmarks and models for generating natural language odor descriptions

ODIEU pairs human odor texts for 10,124 molecules with olfaction-tuned Sentence-BERT evaluators; CIRANO, a frozen GPT-2 with a molecular prefix, generates descriptions approaching human agreement.

Functional Locality-Aligned Learning Reveals Structure-Function Causality in Enzyme Kinetics

This paper introduces EnzymePlex, a framework that improves enzyme kinetic parameter prediction by aligning structural and functional representations, achieving better generalization and mechanistic insights.

EvoStructCLIP: A Mutation-Centered Multimodal Embedding Model for CAGI7 Variant Effect Prediction

EvoStructCLIP predicts missense variant effects by integrating 3D structure and evolutionary data, showing transferability across gene-specific functions and biological tasks.

Contrastive Alignment of Expression and Copy Number Highlights Dosage-Insensitive Genes in Cancer

The paper introduces a contrastive learning framework to identify dosage-insensitive genes in cancer by aligning expression and copy number data from single-cell RNA-seq.

BiGAT-Fusion: Node-Wise Gated Bidirectional Graph Attention for Drug Repurposing

BiGAT-Fusion improves drug-disease association prediction by addressing class imbalance, directional asymmetry, and static feature fusion through a novel graph neural network model.

MAP: A Knowledge-driven Framework for Predicting Single-cell Responses for Unprofiled Drugs

MAP predicts single-cell responses to unprofiled drugs using a knowledge-driven framework that integrates biological knowledge and improves performance over baselines.

Structural and Metabolic Characterization of Ni(I)-inhibitors Provide a Robust Anti-Methanogenicity Scoring System

The paper develops a scoring system to identify effective anti-methanogenic compounds by analyzing their structural and metabolic properties.

Contrastive learning for antibody-antigen sequence-to-specificity prediction

This paper introduces CALM, a model that predicts antibody-antigen binding from sequences using contrastive learning, achieving 7% top-1 retrieval on a test set.

SpaMOAL: A spatial multi-omics graph contrastive learning method for spatial domains identification

SpaMOAL identifies spatial domains by integrating spatial coordinates, histology images, and multi-omics data through graph contrastive learning, outperforming existing methods.

ABaCo: Addressing Heterogeneity Challenges in Metagenomic Data Integration with Adversarial Generative Models

ABaCo integrates metagenomic data from multiple studies, corrects technical heterogeneity, and outperforms existing methods while preserving biological signals.

Learning from All Views: A Multiview Contrastive Framework for Metabolite Annotation

This paper introduces MVP, a multiview contrastive framework for metabolite annotation that improves molecular candidate ranking using multiple data views.

CLADES - Contrastive Learning Augmented DifferEntial Splicing with Orthologous Positive Pairs

The paper introduces CLADES, a contrastive learning method for differential splicing prediction using orthologous positive pairs, achieving strong performance in classification and regression tasks.

Learning heritable multimodal brain representation via contrastive learning

This paper introduces a multimodal contrastive learning framework to derive heritable brain representations from paired T1- and T2-weighted MRIs, improving genetic discovery and revealing shared biological targets.

Unsupervised anomaly detection for tumor delineation in a preclinical model of glioblastoma using CEST MRI

This paper develops an unsupervised anomaly detection method using CEST MRI to improve glioblastoma delineation in rats, achieving better performance than baseline models.

Causal gene regulatory network inference from Perturb-seq via adaptive instrumental variable modeling

ADAPRE infers causal gene regulatory networks from Perturb-seq data by using adaptive instrumental variable modeling, improving accuracy over existing methods.

Transformers Outperform ConvNets for Root Segmentation: A Systematic Comparison Across Nine Datasets

The paper compares Transformer and ConvNet models for root segmentation across nine datasets, finding Transformers significantly outperform ConvNets in accuracy and root-diameter agreement.

Transfer Learning Assessment of Data-driven Crystallisation Processes via Constrained Neural Ordinary Differential Equations

This paper assesses transfer learning for crystallisation modelling using constrained neural ODEs, showing improved generalisation and physical consistency in data-scarce scenarios.

Deciphering DEL Pocket Patterns Through Contrastive Learning

This paper introduces ErePOC, a contrastive learning model for representing protein binding pockets, which improves DEL target selection by identifying shared characteristics and functional similarities.

Enzyme Classification via Semi-Supervised Functional ResidueLearning

This paper introduces SLEEC, a semi-supervised framework for enzyme classification that achieves state-of-the-art performance and provides residue-level annotations.

Improving SCVI for low-count cells through self-supervised augmentation

The paper improves SCVI by addressing low-UMI cell representation issues through self-supervised augmentation, preserving cell identity and variation at lower UMI depths.

CASTLE: Cell-type Aware SpaTial domain detection via contrastive Learning Embedding

CASTLE improves spatial domain detection by integrating cell-type information and contrastive learning, achieving better accuracy across tissues and resolutions.

SpatialESD: spatial ensemble domain detection in spatial transcriptomics

SpatialESD improves spatial domain detection in spatial transcriptomics by integrating multiple methods and capturing both direct and indirect relationships between clusters.

Latent-centric Isotropic Resolution Enhancement for Expansion Microscopy Imaging via Neural Compression and Self-supervised Learning

This paper introduces a self-supervised framework for isotropic super-resolution and compression in expansion microscopy, enabling efficient storage and analysis of whole-organ datasets.

Prediction of bacterial protein-compound interactions with only positive samples

This paper introduces BIN-PU, a PU learning framework for predicting bacterial protein-compound interactions using only positive samples, showing improved performance over existing methods.

DecoderTCR: Compositional Pretraining and Entropy-Guided Decoding for TCR-pMHC Interactions

DecoderTCR improves TCR-pMHC interaction prediction using pretraining and entropy-guided decoding, achieving high AUROC scores without epitope-specific training.

Computational Convergence of Adaptive Immunity and Artificial Intelligence

The paper shows that transformer attention and contrastive learning can be mathematically derived from immunological principles, highlighting computational convergence between adaptive immunity and AI.

Accurate plasmid reconstruction from metagenomics data using assembly-alignment graphs and contrastive learning

PlasMAAG improves plasmid reconstruction from metagenomics data using assembly-alignment graphs and contrastive learning, outperforming existing methods in accuracy and efficiency.