contrastive learning

Full-Length Molecular Models of Brain-Derived α-Synuclein Fibrils Reveal a Fuzzy-Coat-Mediated Mechanism for Selective Peptide Binding

HYPER-Net: Physics-Conditioned Self-Supervised Reconstruction for Fourier Light-Field Microscopy

HYPER-Net enables physics-conditioned self-supervised reconstruction for Fourier light-field microscopy, improving volumetric imaging across diverse biological contexts.

Peering Inside the Black Box: Explainable AI to Interpret Advanced Computer Vision Fungal Pathogen Prediction

The paper trains AI models to identify fungal pathogens from microscopy images and uses explainable AI to reveal the features influencing predictions.

Orthrus: Towards Evolutionary and Functional RNA Foundation Models

Orthrus is a Mamba-based RNA foundation model pre-trained with biological augmentations and contrastive learning, achieving better mRNA property prediction than existing models with less fine-tuning data.

Features fusion or not: harnessing multiple pathological foundation models using Meta-Encoder for downstream tasks fine-tuning

The paper introduces Meta-Encoder, a framework that combines multiple pathological foundation models to improve downstream task performance in computational pathology.

TFBindFormer: A Cross-Attention Transformer for Transcription Factor-DNA Binding Prediction

TFBindFormer improves TF-DNA binding prediction by integrating DNA and TF protein features using cross-attention, outperforming DNA-only models.

Enhancing Biomedical Optical Volumetric Imaging via Self-Supervised Orthogonal Learning

This paper introduces VALID, a self-supervised denoising method for volumetric biomedical imaging that improves structural fidelity across multiple modalities.

TPCAV: Interpreting deep learning genomics models via concept attribution

This paper introduces TPCAV, a method to interpret genomics deep learning models by adapting concept attribution, showing it can analyze both motif features and general genomic concepts.

Molecular Contrastive Learning with Graph Attention Network (MoCL-GAT) for Enhanced Molecular Representation

The paper introduces MoCL-GAT, a self-supervised learning framework for molecular representation, achieving state-of-the-art performance on molecular property prediction tasks.

Protein Language Models Encode Evolutionary Grammar but Conflate Topological and Thermodynamic Phases

This paper shows that protein language models like ESM-2 encode evolutionary grammar rather than three-dimensional geometry, conflate topological and thermodynamic phases, and require explicit physical constraints for accurate modeling.

ADAPTIVE MULTI-SCALE GRAPH TRANSFORMER FRAMEWORK FORHISTOPATHOLOGICAL IMAGES

The paper introduces an Adaptive Multi-Scale Graph Transformer (AMGT) for histopathological image classification, achieving better performance and efficiency than existing models.

BASSL-MI: Batch-Agnostic Self-Supervised Learning Uncovers Clinically Relevant Tumor Niches in Multiplexed Imaging

BASSL-MI uses self-supervised learning to identify clinically relevant tumor niches in multiplexed imaging without requiring cell segmentation.

Wavelet-Domain Multi-Representation and Ensemble Learning for Automated ECG Analysis

This paper uses wavelet transforms and ensemble learning to classify ECGs into five diagnostic classes, achieving an AUC of 0.9233 on the PTB-XL dataset.

From nucleotides to semantics: genomic representation learning via joint-embedding predictive architecture

This paper introduces GenoJEPA, a genomic representation learning framework that improves efficiency and performance in downstream tasks by focusing on semantic alignment rather than nucleotide-level reconstruction.

Methods for Molecular Recognition Computing

The paper establishes a neural architecture for molecular recognition inspired by physics principles, achieving high data efficiency in antibody-antigen retrieval.

Simplex-Constrained Neural Topic VAEs with Flow Refinement for Interpretable Single-Cell Gene-Program Discovery

This paper introduces Topic-FM, a neural topic VAE for single-cell transcriptomics that improves interpretability and clustering performance across multiple datasets.

MoCoO: Momentum Contrast ODE-Regularized VAE for Single-Cell Trajectory Inference and Representation Learning

MoCoO improves single-cell trajectory inference by combining VAE, Neural ODE, MoCo, and Flow Matching, achieving top performance in clustering and embedding quality across multiple datasets.

scTGCL: A Transformer-Based Graph Contrastive Learning Approach for Efficiently Clustering Single-Cell RNA-seq Data

scTGCL improves scRNA-seq clustering by using a Transformer-based graph contrastive learning framework that handles dropout and noise, achieving high accuracy and efficiency.

TSProm: Deciphering the Genomic Context of Tissue Specificity

TSProm identifies tissue-specific gene regulatory elements by training specialized DNA models and using explainable AI to analyze promoter sequences in human tissues.

Natural variation in transplacental transfer efficiency exposes distinct transcriptional network architectures of PFAS effects on birth weight and gestational age

The paper shows PFAS affect birth weight and gestational age through distinct transcriptional network architectures, with network position, not fold-change, determining mediator importance.

Cross-species single-cell atlases chart progression, therapy-driven remodelling and immune evasion in pancreatic cancer

This paper generates cross-species single-cell atlases of pancreatic cancer to study progression, therapy response, and immune evasion across human and mouse models.

Robust Trachea Segmentation from CT Imaging Using Fully Automated and Prompt-Based Models

This paper compares automated and prompt-based models for trachea segmentation in CT scans, showing that dataset structure affects reliability and interpretability.

InSTaPath: Integrating Spatial Transcriptomics and histoPathology Images via Multimodal Topic Learning

InSTaPath integrates spatial transcriptomics and histopathology images using multimodal topic learning to improve spatial domain identification and uncover gene-morphology relationships.

scRGCL: Neighbor-Aware Graph Contrastive Learning for Robust Single-Cell Clustering

scRGCL improves single-cell clustering by integrating contrastive learning with cluster-level guidance, achieving better performance on multiple datasets.

A Unified Dynamical-Systems and Control-Theoretic Model for Single-Cell Fate Dynamics

This paper introduces a unified dynamical-systems and control-theoretic model to predict single-cell fate dynamics by integrating pseudotime, RNA velocity, optimal transport, and stochastic trajectories under partial observability.

Dual-encoder contrastive learning accelerates enzyme discovery

This paper introduces Horizyn-1, a deep learning framework for enzyme discovery that improves performance with larger datasets and validates its effectiveness through experimental testing.

RNAElectra: An ELECTRA-Style RNA Foundation Model for RNA Regulatory Inference

RNAElectra is an RNA foundation model pretrained using replaced-token detection, showing strong performance across RNA regulatory tasks and supporting interpretability.

Vascular smooth muscle cell atherosclerosis trajectories characterized at single cell resolution identify causal transcriptomic and epigenomic mechanisms of disease risk

The paper characterizes atherosclerosis trajectories in vascular smooth muscle cells using single-cell data and identifies causal transcriptomic and epigenomic mechanisms of CAD risk.

VarDCL: A Multimodal PLM-Enhanced Framework for Missense Variant Effect Prediction via Self-distilled Contrastive Learning

VarDCL predicts pathogenic missense variants by integrating multimodal protein language model embeddings and self-distilled contrastive learning, achieving high performance on clinical variants.

PepCABO: Latent-space Bayesian optimization for peptide-MHC binding using contrastive alignment

PepCABO improves peptide-MHC binding optimization using contrastive alignment and Bayesian optimization in a latent space, outperforming baselines across multiple alleles.