variational autoencoder

Modeling Transient Brain Coactivity Patterns in Latent Space with FMRI Data

This study uses a VAE to map fMRI data into a latent space, enabling the modeling of transient co-activity patterns and their temporal dynamics.

Multi-resolution deconvolution of spatial transcriptomics data reveals continuous patterns of inflammation

The paper introduces DestVI, a method for multi-resolution spatial transcriptomics analysis that captures continuous cell state variation, and applies it to study inflammation in lymph nodes and tumors.

Microbiome-based disease prediction with multimodal variational information bottlenecks

This paper introduces MVIB, a deep learning model that combines gut microbiome data with metabolomic data to predict disease, achieving high performance across multiple disease cohorts.

Deciphering spatial domains from spatially resolved transcriptomics with adaptive graph attention auto-encoder

This paper introduces STAGATE, a graph attention auto-encoder for spatial domain detection in spatial transcriptomics, improving accuracy and reducing batch effects across sections.

Hyperbolic Geometry-Based Deep Learning Methods to Produce Population Trees from Genotype Data

This paper introduces deep learning models using hyperbolic geometry to generate population trees from genotype data, showing improved performance and generalizability compared to traditional methods.

High resolution species assignment of Anopheles mosquitoes using k-mer distances on targeted sequences

This paper introduces NNoVAE, a method for high-resolution species assignment of Anopheles mosquitoes using k-mer distances, achieving 98% accuracy for the An. gambiae complex and 89% for other species.

The RNA Virome of Echinoderms

This study characterizes the RNA virome of 38 echinoderm species, identifying 347 viral genome fragments across nine viral orders and finding no significant difference in viral reads between life stages.

Simultaneous dimensionality reduction and integration for single-cell ATAC-seq data using deep learning

This paper introduces a variational auto-encoder for single-cell ATAC-seq data that performs dimensionality reduction and batch correction simultaneously.

ACTIVA: realistic single-cell RNA-seq generation with automatic cell-type identification using introspective variational autoencoders

ACTIVA generates realistic synthetic scRNA-seq data with cell-type identification, improving classifier performance and reducing training time compared to existing methods.

A model of working memory for latent representations

The paper introduces a model of working memory that encodes and retrieves visual knowledge through latent representations, showing efficient encoding of familiar items and storage of novel patterns.

conST: an interpretable multi-modal contrastive learning framework for spatial transcriptomics

conST learns embeddings from gene expression, spatial neighbors, and histology via three-level contrastive learning, reaching adjusted Rand index 0.65 on spatialLIBD slice 151673, with GNNExplainer-based explanations.

Semi-supervised single-cell cross-modality translation using Polarbear

Polarbear translates between single-cell modalities using semi-supervised learning, improving accuracy over existing methods and enabling cross-modality cell matching.

FAME: Fragment-based Conditional Molecular Generation for Phenotypic Drug Discovery

FAME generates novel molecules with desired biological activity by learning from gene expression data using a conditional variational autoencoder and fragment-based generation.

Benchmarking atlas-level data integration in single-cell genomics

BBKNN, Scanorama, and scVI lead complex scRNA-seq integration in a 38-setup, nine-task, 14-metric benchmark; highly variable gene selection helps, whereas scaling favors batch removal over bio-conservation.

Inductive inference of novel protein-molecule interactions using Heterogeneous Graph Transformer (HGT) AutoEncoder

This paper uses a Heterogeneous Graph Transformer AutoEncoder to infer novel protein-molecule interactions by integrating physicochemical and biological information from a heterogeneous graph.

Contrastive Cycle Adversarial Autoencoders for Single-cell Multi-omics Alignment and Integration

This paper introduces a framework to align and integrate single-cell RNA-seq and ATAC-seq data, achieving higher accuracy than existing methods on both simulated and real data.

Modeling the trajectory of SARS-CoV-2 spike protein evolution in continuous latent space using a neural network and Gaussian process

This paper models SARS-CoV-2 spike protein evolution using a neural network and Gaussian process to predict future variants up to five months in advance.

Representation learning of RNA velocity reveals robust cell transitions

This paper introduces VeloAE, a method to improve the accuracy and stability of cell transition estimation from RNA velocity data across multiple datasets.

Learning the rules of cell competition without prior scientific knowledge

This paper develops a deep learning model to learn the rules of cell competition from time-lapse microscopy data and identifies cell density as the key factor in determining cell fate.

Artificial Intelligence Guided Conformational Mining of Intrinsically Disordered Proteins

This paper uses generative autoencoders to predict conformational ensembles of intrinsically disordered proteins from short molecular dynamics simulations.

A fast variational algorithm to detect the clonal copy number substructure of tumors from single-cell data

SCEVAN is a fast variational algorithm that deconvolves clonal substructure in tumors from single-cell data, accurately distinguishing malignant and non-malignant cells.

Photorealistic reconstruction of visual texture from EEG signals

This paper reconstructs photorealistic texture images from EEG signals using an MVAE-based approach, showing that texture perception can be decoded from low-resolution neural data.

Investigating Carboxysome Morphology Dynamics with a Rotationally Invariant Variational Autoencoder

This study uses a rotationally invariant variational autoencoder to analyze carboxysome morphology dynamics in cyanobacteria, finding it can quantify changes in location, shape, and size over time.

Scalable estimation of microbial co-occurrence networks with Variational Autoencoders

This paper introduces a scalable method using Variational Autoencoders to estimate microbial co-occurrence networks from high-dimensional microbiome data.

Time-Dependent Image Restoration of Low-SNR Live Cell Ca2+ Fluorescence Microscopy Data

This paper introduces a variational image restoration method for low-SNR live cell Ca2+ fluorescence microscopy data, demonstrating its effectiveness across synthetic and biological datasets.

Deep learning based on stacked sparse autoencoder applied to viral genome classification of SARS-CoV-2 virus

This paper applies stacked sparse autoencoders to classify SARS-CoV-2 genomes using k-mer image representations, achieving good performance in taxonomic classification.

iLQR-VAE : control-based learning of input-driven dynamics with applications to neural data

The paper introduces iLQR-VAE, a control-based variational inference method for learning dynamics and inputs from neural data, showing effectiveness on synthetic and primate reaching task data.

Spatiotemporal Trajectories in Resting-state FMRI Revealed by Convolutional Variational Autoencoder

This study uses a convolutional variational autoencoder to model spatiotemporal trajectories in resting-state fMRI data, revealing a small number of orthogonal patterns that capture key features of brain activity.

Partitioning variability in animal behavioral videos using semi-supervised variational autoencoders

This paper introduces a video analysis tool that combines supervised pose estimation with unsupervised dimensionality reduction to extract interpretable behavioral features from animal videos, improving downstream analyses.

Predicting chemotherapy response using a variational autoencoder approach

This study uses a variational autoencoder to encode tumor transcriptomes and combines it with XGBoost to predict chemotherapy response across five cancer types, achieving better performance than PCA or raw data in four types.