contrastive learning

Self-supervision advances morphological profiling by unlocking powerful image representations

This study uses self-supervised learning to improve image representation for Cell Painting, outperforming traditional tools in classification tasks and reducing computational costs.

Self-supervised image restoration in coherent X-rayneuronal microscopy

This paper introduces a self-supervised image restoration method for X-ray holographic nanotomography to improve resolution, contrast, and speed in neuronal imaging.

Validation of ten federated learning strategies for multi-contrast image-to-image MRI data synthesis from heterogeneous sources

The paper validates ten federated learning strategies for synthesizing multi-contrast MRI data from heterogeneous sources, finding that FedBAdam outperforms other methods in accuracy and stability.

Identifying patterns differing between high-dimensional datasets with generalized contrastive PCA

This paper introduces gcPCA, a hyperparameter-free method for comparing high-dimensional datasets, demonstrating its utility in detecting hippocampal replay and type II diabetes heterogeneity.

ECLARE: multi-teacher contrastive learning via ensemble distillation for diagonal integration of single-cell multi-omic data

ECLARE integrates unpaired single-cell multi-omic data using multi-teacher contrastive learning, improving cell pairing accuracy and biological structure preservation in depression studies.

CANDI: self-supervised, confidence-aware denoising imputation of genomic data

CANDI improves epigenome imputation by predicting raw counts, handling covariates, and providing uncertainty measures using a self-supervised Transformer model.

Pre-trained Maldi Transformers improve MALDI-TOF MS-based prediction

The paper introduces Maldi Transformer, a pre-trained model for MALDI-TOF MS data that improves prediction tasks through self-supervised learning.

LEVERAGING TRANSFER LEARNING FOR HIGH-ACCURACY PHENOTYPIC SCREENING IN ZEBRAFISH IMAGE ANALYSIS

This paper presents a transfer learning method for classifying zebrafish images before and after drug treatment, achieving high accuracy in detecting phenotypic changes.

A self-supervised deep learning pipeline for segmentation in two-photon fluorescence microscopy

This paper develops a self-supervised deep learning pipeline for TPFM image segmentation, showing that SSL models can match or exceed fully supervised models with less labeled data.

BuDDI: Bulk Deconvolution with Domain Invariance to predict cell-type-specific perturbations from bulk

BuDDI predicts cell-type-specific perturbations from bulk data by integrating single-cell and bulk profiles using domain adaptation in a variational autoencoder.

The reasonable effectiveness of domain adaptation for inference of introgression

This paper demonstrates that domain adaptation improves the accuracy of detecting introgression in population genetics, even when unmodelled processes like ghost introgression are present.

Biologically relevant integration of transcriptomics profiles from cancer cell lines, patient-derived xenografts and clinical tumors using deep learning

MOBER integrates transcriptomic data from cancer cell lines, xenografts, and clinical tumors using deep learning to remove batch effects and improve clinical translatability.

Evaluation of Generative AI Models for Processing Single-Cell RNA-Sequencing Data in Human Pancreatic Tissue

This study evaluates generative AI models for improving cell-type classification in scRNA-seq data by generating synthetic cells and comparing model performance.

Contrastive Learning for Omics-guided Whole-slide Visual Embedding Representation

CLOVER integrates multi-omics data with whole-slide images to generate molecular-aware slide representations, outperforming existing methods in cancer subtype classification and biomarker prediction.

Debiasing Sinkhorn divergence in optimal transport of cellular dynamics

This paper evaluates how Sinkhorn divergence affects optimal transport in single-cell RNA-seq analysis of cellular dynamics, showing its effectiveness depends on temporal sparsity.

Supervised contrastive learning enhances MHC-II peptide binding affinity prediction

This paper introduces ConBoTNet, a deep learning model that improves MHC-II peptide binding affinity prediction using supervised contrastive learning and bottleneck transformers, achieving better performance than existing methods.

Semi-supervised Contrastive Learning for Bioactivity Prediction using Cell Painting Image Data

This paper introduces a semi-supervised contrastive learning method for bioactivity prediction from Cell Painting image data, improving classification of pharmacological classifications and biological targets.

Functional protein mining with conformal guarantees

This paper introduces a method for functional protein mining with statistical guarantees, improving reliability in homology detection and functional annotation.

High-Resolution Time-Lapse Imaging of Droplet-Cell Dynamics via Optimal Transport and Contrastive Learning

This paper develops a method to track microfluidic droplets over time using optimal transport and contrastive learning, achieving high accuracy in tracking without labeled data.

Extracting task-relevant preserved dynamics from contrastive aligned neural recordings

This paper introduces CANDY, a framework that aligns neural and behavioral data across sessions and subjects, enabling cross-session behavior decoding and identifying shared neural dynamics.

Reassessing Deep Learning (and Meta-Learning) Computer Vision as an efficient method to determine taphonomic agency in bone surface modifications

This study reassesses the efficiency of deep learning in taphonomic analysis by replicating and optimizing models to address dataset biases and improve reliability.

Regularized partial correlation provides reliable functional connectivity estimates while correcting for widespread confounding

This paper introduces regularized partial correlation methods to improve functional connectivity estimates in fMRI by reducing noise and confounding, showing higher reliability and accuracy compared to standard methods.

MTGCL: Multi-Task Graph Contrastive Learning for Identifying Cancer Driver Genes from Multi-omics Data

MTGCL identifies cancer driver genes using multi-task graph contrastive learning, improving performance by integrating multi-omics data and unlabeled node information.

TOGGLE delineates fate and function within individual cell types via single cell transcriptomics

TOGGLE identifies functional heterogeneity within cell types using single-cell transcriptomics, revealing distinct states like ferroptosis and apoptosis in ischemic stroke and uncovering epigenetic memory in neural stem cells.

PRSformer: Disease Prediction from Million-Scale Individual Genotypes

PRSformer predicts disease risk from large-scale genotypes using deep learning, outperforming linear models and summary-statistic methods in multi-ancestry cohorts.

Enhancing Structure-aware Protein Language Models with Efficient Fine-tuning for Various Protein Prediction Tasks

This paper introduces S-PLM, a 3D structure-aware protein language model that improves prediction tasks by integrating structural information through efficient fine-tuning.

Sequence-based coevolutionary prediction of species-specific interactomes

This study introduces CM2.0, a coevolution-based method to predict species-specific PPIs, revealing conserved and species-specific interactions in E. coli, S. enterica, and S. aureus.

Delineating the Effective Use of Self-Supervised Learning in Single-Cell Genomics

This study benchmarks self-supervised learning techniques in single-cell genomics, finding masked autoencoders outperform contrastive methods in cell type prediction and data integration.

DNA-binding domain-aware classification enables systematic annotation of the regulatory genome

TFClassPredict is a DNABERT-based framework that classifies DNA-binding domains to annotate regulatory elements, outperforming existing methods and enabling lineage-specific TF program reconstruction.

An unsupervised framework for comparing SARS-CoV-2 protein sequences using LLMs

This paper proposes an unsupervised framework using large language models to compare SARS-CoV-2 protein sequences, showing improved clustering performance for emerging variants.