variational autoencoder

Variational inference accelerates accurate DNA mixture deconvolution

This paper shows that variational inference methods like SVGD and VI can speed up DNA mixture deconvolution with minimal loss of accuracy.

Towards A More Informative Representation of the Fetal-Neonatal Brain Connectome using Variational Autoencoder

This paper uses a variational autoencoder to model non-linear fetal-neonatal brain development, improving age prediction and revealing functional brain networks.

Foldseek: fast and accurate protein structure search

Foldseek's 3Di tertiary-interaction alphabet enables structure search four to five orders of magnitude faster than Dali and TM-align at 86% and 88% of their respective sensitivities.

Representation Learning to Effectively Integrate and Interpret Omics Data

This paper introduces a framework to integrate omics data types and generate interpretable embeddings by addressing regularization challenges in deep generative models.

The Power of Two: integrating deep diffusion models and variational autoencoders for single-cell transcriptomics analysis

This paper introduces scVAEDer, a deep-learning model combining variational autoencoders and diffusion models to generate meaningful embeddings for single-cell transcriptomics data, enabling data generation, perturbation prediction, and regulator detection.

Scalable variational inference for low-rank spatio-temporal receptive fields

The paper introduces a scalable variational inference method for estimating low-rank spatio-temporal receptive fields in neural data, showing improved performance over existing estimators.

Fast and Accurate Bayesian Polygenic Risk Modeling with Variational Inference

VIPRS improves polygenic risk prediction using variational inference, achieving high accuracy and efficiency in both simulated and real datasets.

A scalable variational approach to characterize pleiotropic components across thousands of human diseases and complex traits using GWAS summary statistics

FactorGo identifies pleiotropic components across thousands of human diseases and traits using GWAS summary statistics, outperforming existing methods in capturing latent factors.

rcCAE: a convolutional autoencoder method for detecting intra-tumor heterogeneity and single-cell copy number alterations

rcCAE uses a convolutional autoencoder to detect intra-tumor heterogeneity and single-cell copy number alterations from scDNA-seq data, outperforming existing methods in simulated and real datasets.

Neurospectrum: A Geometric and Topological Deep Learning Framework for Uncovering Spatiotemporal Signatures in Neural Activity

Neurospectrum extracts interpretable spatiotemporal features from neural activity using a geometric and topological deep learning framework, outperforming traditional methods in simulations and fMRI data.

Discovering highly potent antimicrobial peptides with deep generative model HydrAMP

This paper introduces HydrAMP, a deep generative model for creating potent antimicrobial peptides, which outperforms existing methods and validates nine new peptides in lab tests.

Unsupervised discovery of family specific vocal usage in the Mongolian gerbil

This study uses unsupervised machine learning to discover family-specific vocal usage in Mongolian gerbils, revealing stable vocal repertoire preferences and transitions across weeks.

Genome-scale enzymatic reaction prediction by variational graph autoencoders

The paper introduces MPI-VGAE, a framework for predicting metabolite-protein interactions across organisms, showing improved performance in reconstructing metabolic pathways and disease-specific networks.

Learning Optimal White Matter Tract Representations from Tractography using a Deep Generative Model for Population Analyses

This paper uses a ConvVAE to reduce the dimensionality of white matter tractography data, identifies 6 tracts with significant differences between Alzheimer's patients and controls, and preserves distances in a 6-dimensional space.

ProteinVAE: Variational AutoEncoder forTranslational Protein Design

The paper introduces a variational autoencoder for generating synthetic viral vector serotypes with reduced pre-existing immune response, validated through structural and dynamic analysis.

Generative Molecular Design and Experimental Validation of Selective Histamine H1 Inhibitors

The paper develops a generative molecular design platform, ATOM-GMD, to create selective histamine H1 inhibitors, which are validated experimentally with six compounds showing high potency and selectivity.

Inferring the basis of binaural detection with a modified autoencoder

This study uses modified autoencoders to infer latent mechanisms of binaural detection, finding artificial neurons that mimic cross-correlation dynamics similar to animal models.

Biophysical modeling with variational autoencoders for bimodal, single-cell RNA sequencing data

The paper introduces biVI, a variational autoencoder for bimodal single-cell RNA sequencing data, which models biophysical processes to capture cell type structure and parameter values.

Active Learning of the Conformational Ensemble of Proteins using Maximum Entropy VAMPNets

This paper introduces MaxEnt VAMPNet, an active learning method that accelerates conformational landscape exploration in biomolecules by maximizing entropy in VAMPNet-trained microstates.

Variational graph encoders: a surprisingly effective generalist algorithm for holistic computer-aided drug design

The paper introduces a generalist model for drug design that predicts multiple properties simultaneously, achieving faster virtual screening and drug development with explainability.

A Convolutional Autoencoder-based Explainable Clustering Approach for Resting-State EEG Analysis

This paper introduces a convolutional autoencoder-based clustering method for resting-state EEG data, identifying 8 EEG states in schizophrenia and linking time outside the dominant state to negative symptom severity.

Hierarchical VAEs provide a normative account of motion processing in the primate brain

This study shows hierarchical VAEs better model motion perception by aligning with brain function and ecological stimulus statistics.

scTopoGAN: Unsupervised manifold alignment of single-cell data

scTopoGAN aligns single-cell datasets with non-overlapping cells or features using topology-guided GANs, outperforming existing methods in unsupervised settings.

Multi-scale tissue fluorescence mapping with fibre optic ultraviolet excitation and generative modelling

This paper introduces FUSE, a fibre optic microscopy technique with UV excitation and generative modelling for high-throughput cellular imaging of thick tissues, achieving sub-nuclear clarity and 3D topography.

Prediction of designer-recombinases for DNA editing with generative deep learning

The paper introduces RecGen, a deep learning model that generates designer-recombinases for novel DNA target sites, outperforming traditional directed evolution methods.

Deep learning-enabled design of synthetic orthologs of a signaling protein

This study uses deep generative models to design synthetic SH3 domains functional in yeast stress signaling, showing a VAE can recapitulate natural function and classify orthologs by phylogeny and function.

De novo design of high-affinity protein binders to bioactive helical peptides

This paper designs high-affinity protein binders for bioactive helical peptides using deep learning and generative models, achieving picomolar affinity without experimental optimization.

Phylogenetic inference using Generative Adversarial Networks

This paper introduces phyloGAN, a GAN for phylogenetic inference, and evaluates its performance on small-scale datasets.

Learning orientation-invariant representations enables accurate and robust morphologic profiling of cells and organelles

This paper introduces O2-VAE, an orientation-invariant representation learning method for morphologic profiling of cells and organelles, improving clustering and outlier detection.

Robust probabilistic modeling for single-cell multimodal mosaic integration and imputation via scVAEIT

This paper introduces scVAEIT, a probabilistic variational autoencoder for integrating and imputing multimodal single-cell data with mosaic measurements, showing improved performance across unseen cell types and tissues.