variational autoencoder

Neural Mechanisms Linking Global Maps to First-Person Perspectives

This paper develops a VAE model with RNNs to study neural mechanisms of perspective transformation between first-person and global map views, finding that temporal sequences and proximal cues are critical.

CoupleVAE: coupled variational autoencoders for predicting perturbational single-cell RNA sequencing data

CoupleVAE predicts post-perturbation single-cell RNA-seq data by using coupled variational autoencoders, achieving better accuracy than existing models in three real datasets.

Unifying multimodal single-cell data with a mixture-of-experts β-variational autoencoder framework

UniVI integrates multimodal single-cell data using a mixture-of-experts β-variational autoencoder, improving cross-modal alignment and label transfer across various data types and study designs.

Local graph-motif features improve gene interaction network prediction

This paper introduces LAMP features to enhance VGAEs for gene interaction network prediction, showing improved performance on missing-link prediction tasks.

Modeling conditional distributions of neural and behavioral data with masked variational autoencoders

The paper introduces a VAE-based method to model conditional distributions between neural and behavioral data, validated on fly and monkey datasets.

Tahoe-100M: A Giga-Scale Single-Cell Perturbation Atlas for Context-Dependent Gene Function and Cellular Modeling

A 100-million-cell atlas: 50 cancer lines pooled as spheroid villages, treated with 379 drugs at several doses, profiled by combinatorial-barcoding scRNA-seq and embedded with a 10-dimensional scVI model.

stDyer enables spatial domain clustering with dynamic graph embedding

stDyer improves spatial domain clustering in SRT data by combining GMVAE and GATs with dynamic graphs and mini-batch sampling, outperforming existing methods.

Nearest-neighbor nonnegative spatial factorization to study spatial and temporal transcriptomics

The paper introduces NNNSF, a scalable variant of NSF using VNNGPs, reducing computational complexity and improving reconstruction loss on spatial and temporal transcriptomics data.

CVAE-based Causal Representation Learning from Retinal Fundus Images for Age Related Macular Degeneration(AMD) Prediction

This paper develops a CVAE-based framework to learn causal representations from retinal fundus images for AMD prediction, achieving better causal disentanglement and reliable diagnosis.

Predicting Mutation-Induced Relative Protein-Ligand Binding Affinity Changes via Conformational Sampling and Diversity Integration with Subsampled Alphafold2 in Few-Shot Learning

The paper predicts mutation-induced changes in protein-ligand binding affinity using few-shot learning with AlphaFold2 subsampling and Siamese learning, achieving improved correlation on TKI datasets.

Multimodal Predictive Modeling: Scalable Imaging Informed Approaches to Predict Future Brain Health

This paper introduces an image-assisted predictive modeling approach that integrates neuroimaging and behavioral data to predict future brain health constructs.

Transformer Graph Variational Autoencoder for Generative Molecular Design

This paper introduces TGVAE, a Transformer Graph Variational Autoencoder that generates diverse and chemically valid molecules by using molecular graphs instead of SMILES strings, outperforming existing methods.

sciLaMA: A Single-Cell Representation Learning Framework to Leverage Prior Knowledge from Large Language Models

sciLaMA integrates gene embeddings from large language models with scRNA-seq data using a paired-VAE architecture, improving downstream tasks like batch correction and clustering while maintaining computational efficiency.

scGRIP: a graph-based explainable AI framework for single-cell multi-omics Gene Regulatory Inference with Prior Knowledge

scGRIP improves gene regulatory network inference in single-cell multi-omics data by integrating prior knowledge and explainable AI techniques.

Inferring single-cell dynamics of a molecular reporter from unlabeled live-cell light microscopy data analyzed with delay embedding

The paper uses delay embedding and VAE features to predict single-cell dynamics of a molecular reporter from unlabeled live-cell imaging 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.

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.

Discovery of molecular glues by modeling ternary complex conformational ensembles and thermodynamic stability

This paper introduces GlueMap, a computational platform for discovering molecular glues by integrating structural modeling and machine learning, showing improved performance over existing models in predicting degradation potency and identifying new degraders.

Predicting cell morphological responses to perturbations using generative modeling

This paper introduces IMPA, a generative model that predicts cell morphological responses to chemical and genetic perturbations using untreated cell images as input, showing improved performance on structurally related compounds.

Deep topic modeling of spatial transcriptomics in the rheumatoid arthritis synovium identifies distinct classes of ectopic lymphoid structures

The paper uses spatial transcriptomics and a new method called DeepTopics to identify distinct cellular patterns in RA synovium, revealing divergent ectopic lymphoid structures.

An Enhanced Conditional Variational Autoencoder-Based Normative Model for Neuroimaging Analysis

This paper introduces an enhanced cVAE-based normative model for neuroimaging data, showing comparable performance to existing methods while capturing individual deviations related to hypertension severity.

MODE: high-resolution digital dissociation with deep multimodal autoencoder

MODE improves deconvolution by integrating multi-omic data through a multimodal autoencoder, outperforming existing methods in simulation and real data.

Integration and querying of multimodal single-cell data with PoE-VAE

This paper introduces a VAE-based model for integrating multimodal single-cell data, showing it can map queries across modalities and improve data imputation and biological discovery.

Deep-learning based Embedding of Functional Connectivity Profiles for Precision Functional Mapping

This study uses a variational autoencoder to embed functional connectivity profiles into a latent space for simultaneous comparison across anatomical locations and groups.

EM-Compressor: Electron Microscopy Image Compression in Connectomics with Variational Autoencoders

EM-Compressor uses Variational Autoencoders to compress EM images for connectomics, reducing data size to 1/128th while preserving features needed for segmentation.

Explainable Generative AI Uncovers a Molecular Continuum in Medulloblastoma with Implications for Rare Cancer Subtyping and Treatment Equity

This study uses synthetic data generation and explainable AI to uncover an intermediate subgroup in medulloblastoma, highlighting treatment disparities and supporting the continuum hypothesis between Groups 3 and 4.

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.

Integrating single-cell multimodal epigenomic data using 1D-convolutional neural networks

The paper introduces ConvNet-VAEs, a framework for integrating single-cell multimodal epigenomic data using 1D-convolutional variational autoencoders, showing improved dimension reduction and batch correction.

Using deep generative models for simultaneous representational and predictive modeling of brain and behavior: A graded supervised-to-unsupervised modeling framework.

This paper develops a framework combining generative and discriminative models to evaluate brain-behavior mappings, showing how it captures diverse relationships and identifies assumption violations.