contrastive learning

Multimodal AI Decodes Extreme Environment Functional Dark Matter Beyond Homology

The paper develops ACCESS, a multimodal graph neural network to annotate functional dark matter proteins from extreme environments beyond homology-based methods.

Contrastive Learning Enables Epitope Overlap Predictions for Targeted Antibody Discovery

This paper develops computational methods to predict epitope overlaps in antibodies, showing improved accuracy using contrastive learning and demonstrating their utility in antibody discovery.

Integrating Millions of Years of Evolutionary Information into Protein Structure Models for Function Prediction

This paper introduces ESMSCOP, a contrast-aware pre-training framework that integrates evolutionary and structural information for improved protein function prediction.

Scalable spike sorting across thousands of neurons by modeling neural dynamics with NeuroSort

NeuroSort improves spike sorting by modeling neural dynamics, enabling efficient isolation of thousands of neurons across various recording conditions.

CryoSiam: self-supervised representation learning for automated analysis of cryo-electron tomograms

CryoSiam is a self-supervised framework for cryo-ET data analysis that learns hierarchical representations and transfers to experimental data for denoising, segmentation, and macromolecular detection.

BarlowTrack: A Self-Supervised Framework for Zero-Shot Multi-Object Cell Tracking

This paper introduces BarlowTrack, a self-supervised cell tracking framework that achieves zero-shot accuracy across conditions and labs for neuronal time series in C. elegans.

PyrMol: A Knowledge-Structured Pyramid Graph Framework forGeneralizable Molecular Property Prediction

PyrMol improves molecular property prediction by integrating expert knowledge into a hierarchical graph framework, outperforming existing methods on benchmark datasets.

Lightweight open-source fine-tuning of SAM2 enables domain-specific microscopy segmentation

This paper introduces a lightweight, open-source method to fine-tune SAM2 for microscopy segmentation, achieving high accuracy with small datasets and cross-domain generalization.

Nona: A unifying multimodal masking framework for functional genomics

Nona is a multimodal masking framework that unifies sequence-to-function models, language models, and generative models for functional genomics, improving predictions and revealing privacy vulnerabilities in ATAC-seq data.

Predictive Coding algorithms induce brain-like responses in Artificial Neural Networks

This paper shows that predictive coding-inspired neural networks exhibit brain-like behaviors such as mismatch responses and prior formation, suggesting they can model biological neural networks.

Low rank adaptation of chemical foundation models generate effective odorant representations

The paper shows that fine-tuning chemical foundation models improves odorant-receptor binding prediction, introducing LORAX which outperforms existing models.

BudFinder: A Masked Auto-Encoder Vision Transformer Framework for Yeast Budding Detection

BudFinder uses a masked auto-encoder to detect yeast cell divisions with minimal training data, reducing manual labeling by over five-fold while maintaining accuracy.

MotifAE Reveals Functional Sequence Patterns from Protein Language Model: Unsupervised Discovery and Interpretability Analysis

MotifAE discovers functional sequence patterns from protein language models through an unsupervised framework, improving motif identification and enabling stability prediction.

Atacformer: A transformer-based foundation model for analysis and interpretation of ATAC-seq data

Atacformer is a transformer-based model for scATAC-seq data analysis that improves clustering, cell-type annotation, and cross-modal alignment with RNA-seq data.

FADVI: disentangled representation learning for robust integration of single-cell and spatial omics data

FADVI integrates single-cell and spatial omics data by disentangling batch effects from biological signals, outperforming existing methods across multiple datasets.

Physics-Diffusion-Driven Multiscale Aggregation for Drug-Target Interaction Prediction

PDDMA-DTI improves drug-target interaction prediction by combining physics-inspired diffusion with multiscale aggregation and interaction enhancement, achieving better accuracy and robustness than existing methods.

AnnotateAnyCell: Open-Source AI Framework for Efficient Annotation in Digital Pathology

AnnotateAnyCell reduces annotation time for histopathological images by 25% using a semi-supervised framework with active contrastive learning and human-in-the-loop feedback.

Large mRNA language foundation modeling with NUWA for unified sequence perception and generation

This paper introduces NUWA, an mRNA language model that enables unified sequence perception and generation across diverse species and tasks.

Decoding heart failure subtypes with neural networks via differential explanation analysis

This study identifies differentially explained genes in heart failure subtypes using neural networks and XAI tools to improve understanding of molecular mechanisms.

Anomaly detection for high-content image-based phenotypic cell profiling

This paper introduces an anomaly-based representation for high-content image-based phenotypic cell profiling, improving reproducibility and classification while maintaining interpretability.

Single-cell multiomics data integration and generation with scPairing

scPairing generates multiomics data from single cells using deep learning, capturing biological structures and enabling trimodal data creation.

REPTRA: Mapping Immune T Cell Receptor Activity from Full Sequences with a Debiased Contrastive Loss

The paper introduces REPTRA, a contrastive learning model for TCR-epitope recognition using full TCR sequences and a debiased loss function, showing improved performance and highlighting the importance of CDR1 and CDR2.

NyxBind: enhancing DNN representations via contrastive learning for TFBS prediction

NyxBind improves TFBS prediction by using contrastive learning to enhance DNN representations, outperforming existing models across benchmarks.

Self-supervised learning enables robust microbiome predictions in data-limited and cross-cohort settings

This paper develops self-supervised learning models for gut microbiome metagenomic data, showing improved prediction performance in data-limited and cross-cohort settings.

ASCENT: Annotation-free Self-supervised Contrastive Embeddings for 3D Neuron Tracking in Fluorescence Microscopy

ASCENT tracks neurons in 3D microscopy data without manual annotation, achieving state-of-the-art accuracy on C. elegans datasets.

Icecream: High-Fidelity Equivariant Cryo-Electron Tomography

Icecream improves cryo-ET reconstruction by addressing denoising, missing-wedge recovery, and computational efficiency using equivariant deep learning.

Odyssey: reconstructing evolution through emergent consensus in the global proteome

Odyssey generates and designs proteins using a multimodal language model with a consensus mechanism, achieving strong performance in protein generation and structure discretization.

ECloudGen: Leveraging Electron Clouds as a Latent Variable to Scale Up Structure-based Molecular Design

ECloudGen improves molecular design by using electron clouds as latent variables, generating more potent binders and exploring broader chemical space than existing methods.

scConcept: Contrastive pretraining for technology-agnostic single-cell representations beyond reconstruction

scConcept improves single-cell representations by using contrastive learning instead of reconstruction, outperforming existing methods in multiple tasks.

Mapping the diverse topologies of protein-protein interaction fitness landscapes

The paper identifies target-specific fitness landscapes in protein-protein interactions and develops a supervised contrastive learning approach to predict binders.