contrastive learning

Maximum Classifier Discrepancy Generative Adversarial Network for Jointly Harmonizing Scanner Effects and Improving Reproducibility of Downstream Tasks

This paper introduces MCD-GAN, a method that harmonizes scanner effects in MRI data while preserving biological information, improving cross-scanner classification performance.

MatchCLOT: Single-Cell Modality Matching with Contrastive Learning and Optimal Transport

MatchCLOT improves modality matching in single-cell data by combining contrastive learning and optimal transport, achieving state-of-the-art results on a benchmark challenge.

Digitally Predicting Protein Localization and Manipulating Protein Activity in Fluorescence Images Using Four-dimensional Reslicing GAN

The paper introduces a GAN-based method to predict protein localization in fluorescence images and manipulate protein activity to study localization-based functions.

Validation of predicted conformal intervals for prediction of human clinical pharmacokinetics

The paper validates conformal prediction intervals for human pharmacokinetic parameters using ANDROMEDA software, showing 78% of observed data within 70% confidence intervals.

ConstrastivePose: A contrastive learning approach for self-supervised feature engineering for pose estimation and behavorial classification of interacting animals

This paper introduces ConstrastivePose, a contrastive learning method for self-supervised feature engineering in pose estimation and behavioral classification of interacting animals, showing improved performance over handcrafted features.

Contrasting drugs from decoys

This paper introduces CON-Plex, a drug-target interaction model that improves specificity by distinguishing drugs from decoys using contrastive learning.

Bio-friendly long-term subcellular dynamic recording by self-supervised image enhancement microscopy

This paper introduces DeepSeMi, a self-supervised denoising framework that enhances fluorescence microscopy by increasing SNR and enabling long-term, high-resolution imaging of cellular processes.

Contrastive learning of protein representations with graph neural networks for structural and functional annotations

PenLight improves protein annotation by integrating 3D structure and sequence data using graph neural networks and contrastive learning.

Prediction of Electric Fields Induced by Transcranial Magnetic Stimulation in the Brain using a Deep Encoder-Decoder Convolutional Neural Network

This paper develops a deep convolutional neural network to predict induced electric fields from TMS in real-time using MRI data, achieving high accuracy with peak signal-to-noise ratios of 32.83dB and 28.01dB.

SalienceNet: an unsupervised Image-to-Image translation method for nuclei saliency enhancement in microscopy images

SalienceNet enhances nuclei saliency in low-light microscopy images, improving segmentation accuracy by 30% with Otsu thresholding and 26% with StarDist.

ICAN: interpretable cross-attention network for identifying drug and target protein interactions

The paper introduces ICAN, an attention-based method for predicting drug-target interactions using SMILES and amino acid sequences, achieving high performance and interpretability.

Solving Anscombe's Quartet using a Transfer Learning Approach

The paper proposes ICLUST, a transfer learning approach to visually distinguish scatterplots with similar summary statistics, identifying meaningful clusters in simulations and real data.

Bioacoustic Event Detection with Self-SupervisedContrastive Learning

This paper develops a self-supervised deep learning method for unsupervised bioacoustic event detection, achieving high accuracy in identifying sperm whale clicks and other vocalizations without manual annotation.

Rates of convergence in the two-island andisolation-with-migration models

The paper analyzes the statistical performance of demographic inference methods under continuous migration using theoretical bounds.

Deciphering Spatial Domains by Integrating Histopathological Image and Transcriptomics via Contrastive Learning

This paper introduces ConGI, a method that integrates gene expression and histopathological images via contrastive learning to improve spatial domain identification in spatial transcriptomics.

Natural language supervision with a large and diverse dataset builds better models of human high-level visual cortex

The paper shows that using natural language supervision with large, diverse datasets improves models' ability to predict human high-level visual cortex responses compared to image-only or text-only models.

sciCAN: Single-cell chromatin accessibility and gene expression data integration via Cycle-consistent Adversarial Network

sciCAN integrates single-cell chromatin accessibility and gene expression data using an adversarial network, outperforming existing methods in cross-modal transfer and preserving biological structure in hematopoietic data.

Single-Cell Multi-Modal GAN (scMMGAN) reveals spatial patterns in single-cell data from triple negative breast cancer

scMMGAN integrates multi-omic data to reveal spatial patterns in triple negative breast cancer, showing improved alignment and novel gene associations.

Coupling cryo-electron tomography with mixed-scale dense neural networks reveals re-organization of the invasion machinery of Toxoplasma gondii upon ionophore-stimulation

The paper uses cryo-electron tomography and mixed-scale dense neural networks to analyze the reorganization of Toxoplasma gondii's invasion machinery upon ionophore stimulation, revealing structural changes and evolutionary conservation with Plasmodium falciparum.

Data augmentation for imbalanced blood cell image classification

This paper proposes a method to improve classification of imbalanced blood cell images using data augmentation techniques and a minority class sampling strategy, achieving better performance on minority classes.

Reconstructing Kinetic Models for Dynamical Studies of Metabolism using Generative Adversarial Networks

This paper introduces REKINDLE, a deep learning framework that generates kinetic models of metabolic networks with dynamic properties, demonstrating its efficiency and versatility through three studies.

Self-supervised contrastive learning for integrative single cell RNA-seq data analysis

CLEAR is a self-supervised contrastive learning framework for scRNA-seq that improves clustering, batch correction, and pseudo-time inference by reducing technical noise and heterogeneity.

Attention-aware contrastive learning for predicting T cell receptor-antigen binding specificity

This paper introduces ATMTCR, a contrastive learning model for predicting TCR-antigen binding specificity, showing improved performance over existing methods.

The rubber tree kinome: genome-wide characterization and insights into coexpression patterns associated with abiotic stress responses

This study characterizes the kinome of Hevea brasiliensis, identifying 2,842 PK genes and analyzing their expression and coexpression patterns under abiotic stress conditions.

Cross-Modality and Self-Supervised Protein Embedding for Compound-Protein Affinity and Contact Prediction

This paper introduces cross-modality and self-supervised learning for protein embeddings to improve compound-protein affinity and contact prediction.

Dynamic applicability domain (dAD) for compound-target binding affinity prediction with confidence guarantees

The paper introduces a dynamic applicability domain framework for compound-target binding affinity prediction with confidence guarantees, showing improved performance over existing methods.

Scalable batch-correction method for integrating large-scale single-cell transcriptomes

This paper introduces a scalable batch-correction method for integrating large-scale single-cell transcriptomes and shows it matches state-of-the-art performance on 18 million cells from the Human Cell Atlas.

Semi-Self-Supervised Learning for Semantic Segmentation in Images with Dense Patterns

The paper proposes a semi-self-supervised learning method for wheat head segmentation using minimal manual annotations and achieves high Dice scores on both internal and external datasets.

Attention-wise masked graph contrastive learning for predicting molecular property

This paper introduces a self-supervised learning method, ATMOL, for molecular property prediction using attention-wise graph masking and contrastive learning, achieving state-of-the-art performance.

A Comparison of Methods to Harmonize Cortical Thickness Measurements Across Scanners and Sites

This paper compares four methods for harmonizing cortical thickness data across multiple sites and finds that ComBat-GAM is most effective in detecting case-control differences and age-related cortical thinning.