variational autoencoder

QOBRA: A Quantum Operator-Based Autoencoder for De Novo Molecular Design

QOBRA is a quantum autoencoder for de novo molecular design that generates valid, novel molecules from limited datasets, demonstrating the potential of quantum computing in molecular synthesis.

SurfNet: Reconstruction of Cortical Surfaces via Coupled Diffeomorphic Deformations

SurfNet jointly reconstructs cortical surfaces from MRIs using diffeomorphic deformations, achieving state-of-the-art accuracy and topological correctness.

A Variational Autoencoder Model for Clustering of Cell Nuclei on Microgroove Substrates: Potential for Disease Diagnosis

This study uses a variational autoencoder to cluster nuclei on microgroove substrates, distinguishing between wildtype and laminopathy myoblasts based on nuclear deformation.

MORPH Predicts the Single-Cell Outcome of Genetic Perturbations Across Conditions and Data Modalities

MORPH predicts cellular responses to genetic perturbations across data modalities and conditions using a variational autoencoder and attention mechanism.

Morphology-Aware Profiling of Highly Multiplexed Tissue Images using Variational Autoencoders

This paper introduces a method that combines autoencoders with segmentation to improve single-cell analysis of multiplexed tissue images by capturing morphology and reducing signal spillover.

Elucidating Protein Dynamics through the Optimal Annealing of Variational Autoencoders

This paper introduces an optimal annealing method for VAEs to improve protein conformational dynamics modeling by selecting the optimal β parameter using FVE scores.

Linear scaling reveals low-dimensional structure in observable microbial dynamics

The paper shows that microbial community dynamics can be represented by low-dimensional latent variables, with the critical dimension scaling linearly with observables.

Flight trajectory modeling reveals species-specific obstacle avoidance policies in echolocating bats

This study uses a variational recurrent neural network to show that bat flight trajectories reflect consistent internal policies, revealing structured control strategies in obstacle avoidance.

Providing context: Extracting non-linear and dynamic temporal motifs from brain activity

This paper introduces a deep learning model to analyze resting-state fMRI dynamics, finding that context embeddings better predict schizophrenia diagnosis and correlate with age and symptom severity.

Generative prediction of causal gene sets responsible for complex traits

The paper develops TWAVE, a generative model that identifies causal gene sets for complex traits by integrating transcriptomic data and gene perturbation responses.

Biologically Informed Variational Inference Enables Interpretable Cell Phenotyping and Discovery

NetworkVI integrates multi-omics data using a biologically informed variational autoencoder, enabling interpretable cell phenotyping and discovery of lineage-specific immune states.

Integrating multi-covariate disentanglement with counterfactual analysis on synthetic data enables cell type discovery and counterfactual predictions.

CellDISECT improves cell type discovery and counterfactual predictions by disentangling covariates and using synthetic data for training.

Kaizen: Decomposing cellular images with VQ-VAE

Kaizen improves cell segmentation in microscopy images by using a VQ-VAE to iteratively refine object hypotheses.

Encoding Cell Phenotype from Label-Free Imaging Flow Cytometry with Unsupervised Deep Learning

This study uses a Variational Auto-Encoder to predict microalgal lipid content from label-free Imaging Flow Cytometry data, demonstrating a low-cost alternative to traditional methods.

Deep representation learning for temporal inference in cancer omics: a systematic review

This paper reviews the use of deep representation learning, particularly VAEs, in cancer omics to model temporal processes, highlighting limitations in capturing cancer progression dynamics.

Cross-species and tissue imputation of species-level DNA methylation samples across mammalian species.

The paper develops CMImpute, a method to impute DNA methylation data for unprofiled species-tissue combinations, achieving strong correlation with observed values and imputing 19,786 new combinations.

BrainVAE: Exploring the role of white matter BOLD in preclinical Alzheimer disease classification

This paper introduces BrainVAE, a transformer-based variational autoencoder, to classify preclinical Alzheimer's disease using white matter BOLD signals and finds it outperforms other models when combining white and gray matter data.

Design of diverse, functional mitochondrial targeting sequences across eukaryotic organisms using variational autoencoder

This paper uses a variational autoencoder to design novel mitochondrial targeting sequences across eukaryotic organisms, showing they are functional and diverse.

A joint probabilistic model of human scene and object recognition via non-hierarchical residual computation

The paper introduces a joint probabilistic model for human scene and object recognition using residual computation, showing functional specialization in coarse-to-fine processing.

Dynamic Differences in Clinically relevant Pen β-Lactamases from Burkholderia spp.

This study uses computational methods to analyze dynamic differences among Pen β-lactamases from Burkholderia spp., identifying critical mutations and functional hotspots.

Evaluating discrepancies in dimensionality reduction for time-series single-cell RNA-sequencing data

The paper evaluates discrepancies in dimensionality reduction techniques for time-series scRNA-seq data using synthetic data and finds that different methods may not reliably represent dynamics when used alone.

Protocol for constructing correlation-based molecular networks from large-scale untargeted metabolomics data

This paper describes a protocol using MetVAE to construct molecular networks from metabolomics data, identifying lipid correlations in a mouse model.

ARTEMIS integrates autoencoders and schrodinger bridges to predict continuous dynamics of gene expression, cell population and perturbation from time-series single-cell data

Disentanglement of batch effects and biological signals across conditions in the single-cell transcriptome

This paper introduces Kanade, a variational autoencoder-based method that disentangles batch effects from biological signals in single-cell transcriptomics, improving comparative analysis across conditions.

Improving single molecule localisation microscopy reconstruction by extending the temporal context

The paper improves single-molecule localization microscopy by extending temporal context using a hybrid U-Net and Transformer model, achieving better resolution than existing methods.

Partially characterized topology guides reliable anchor-free scRNA integration

The paper introduces scCRAFT, an autoencoder for scRNA-seq batch integration that outperforms existing methods by preserving biological heterogeneity and handling unbalanced batches and rare cell types.

Protein Sequence Modelling with Bayesian Flow Networks

This paper introduces ProtBFN, a Bayesian Flow Network for protein sequence generation, which outperforms existing models and is fine-tuned for antibody generation.

Gradient matching accelerates mixed-effects inferencefor biochemical networks

The paper introduces GMGTS, a method that accelerates mixed-effects inference for biochemical networks by replacing numerical integration with gradient matching, improving computational efficiency.

Imputing missing minimum inhibitory concentration (MIC) values for Pseudomonas aeruginosa strains with a Denoising AutoEncoder

The paper introduces MICFiller, a Denoising AutoEncoder to impute missing MIC values for Pseudomonas aeruginosa, achieving high accuracy compared to existing methods.

SpaBatch: Batch Alignment of Spatial Transcriptomics Data using Graph Deep Learning

SpaBatch aligns spatial transcriptomics data across batches using graph deep learning to reduce technical variability and enable multi-slice integration.