variational autoencoder

Latent-based Directed Evolution accelerated by Gradient Ascent for Protein Sequence Design

This paper introduces LDE, a variational autoencoder-based method for protein sequence design that improves efficiency by prioritizing high-fitness mutants in latent space, outperforming baselines on eight tasks.

Enhanced cell segmentation with limited annotated data using generative adversarial networks

This paper proposes a customized CycleGAN to enhance cell segmentation with limited annotated data, showing improved performance over conventional methods.

The covariance environment defines cellular niches for spatial inference

This paper introduces COVET and ENVI to model spatial niches and improve gene expression imputation and spatial context inference from single-cell data.

Holographic-(V)AE: an end-to-end SO(3)-Equivariant (Variational) Autoencoder in Fourier Space

This paper introduces H-(V)AE, an SO(3)-equivariant autoencoder in Fourier space for unsupervised learning and generation of 3D data, showing it can encode spherical images and predict protein-ligand binding affinity.

Inferring gene regulatory networks by hypergraph variational autoencoder

This paper introduces HyperG-VAE, a Bayesian deep generative model for GRN inference using scRNA-seq data, showing effectiveness in clustering, visualization, and revealing gene regulation patterns.

Biologically Interpretable VAE with Supervision for Transcriptomics Data Under Ordinal Perturbations

The paper introduces EXPORT, a VAE framework for analyzing ordinally perturbed transcriptomics data, enhancing interpretability by incorporating pathway knowledge and ordinal constraints.

CLigOpt: Controllable Ligand Design through Target-Specific Optimisation

CLigOpt generates target-specific ligands with desired properties using a controllable fragment-based approach, improving the proportion of feasible active molecules by 10%.

Insights into Cellular Evolution: Temporal Deep Learning Models and Analysis for Cell Image Classification

This study compares temporal deep learning models for classifying sequences of cell colony images to understand cellular state transitions.

Generative Adversarial Implicit Successor Representation

This paper introduces a method to implicitly encode Successor Representations using a GAN, enabling action decisions without explicit probability calculations.

Unsupervised generative AI discovers pan-leukocyte dysregulated pathways in single-cell lupus data

The paper applies a Variational Auto Encoder to single-cell RNA-seq data to identify dysregulated pathways in lupus, finding new insights without requiring expert user input.

Quantitative image analysis pipeline for detecting circulating hybrid cells in immunofluorescence images with human-level accuracy

The paper presents an image analysis pipeline for detecting circulating hybrid cells in immunofluorescence images with human-level accuracy using a β-VAE and SVM classifier.

Learning site-invariant features of connectomes to harmonize complex network measures

The paper develops a variational autoencoder to harmonize multi-site diffusion MRI connectomes by isolating site-invariant features and preserving age-related modularity correlations.

VICTree - a Variational Inference method for Clonal Tree reconstruction

VICTree is a variational inference method for clonal tree reconstruction that uses a Tree-structured Mixture Hidden Markov Model and evaluates on cancer datasets.

Identification of Mitosis Stages Using Artificial Neural Networks for 3D Time Lapse Cell Sequences

This study uses 3D CNN models to classify mitosis stages in 3D time-lapse cell sequences, achieving higher accuracy than previous methods.

Inferring the Joint Distribution of Structural and Functional Connectivity in the Human Brain using UNIT-DDPM

This paper introduces UNIT-DDPM, a method to infer the joint distribution of structural and functional connectomes in the human brain, showing improved prediction accuracy and structure-function coupling.

A universal and effective variational method for destriping: application to light-sheet microscopy, FIB-SEM and remote sensing images

This paper introduces a variational method for removing stripe artifacts from images, showing effectiveness in light-sheet microscopy, FIB-SEM, and remote sensing data.

A Generative Model of Memory Construction and Consolidation

This paper presents a computational model of memory construction and consolidation, showing how hippocampal replay trains neocortical generative models to reconstruct sensory experiences and explain memory distortions.

Neural Network-Assisted Humanization of COVID-19 Hamster scRNAseq Data Reveals Matching Severity States in Human Disease

This paper uses neural networks to match disease states between human and hamster COVID-19 scRNAseq data, finding similar cellular responses in specific cell types across species.

FAVA: High-quality functional association networks inferred from scRNA-seq and proteomics data

FAVA infers high-quality functional protein interaction networks from scRNA-seq and proteomics data, outperforming existing methods and identifying interactions for understudied proteins.

A Supervised Contrastive Framework for Learning Disentangled Representations of Cell Perturbation Data

This paper introduces SC-VAE, a supervised contrastive framework for analyzing cell perturbation data, showing improved sensitivity in detecting perturbation effects compared to existing methods.

scTIE: data integration and inference of gene regulation using single-cell temporal multimodal data

scTIE integrates temporal multimodal data and infers gene regulatory relationships predictive of cellular state changes, outperforming existing methods in data integration and biological signal preservation.

OmniClustifyXMBD: Uncover putative cell stateswithin multiple single-cell omics datasets

OmniClustifyXMBD identifies putative cell states across diverse single-cell omics datasets by integrating adaptive signal isolation with a deep variational Gaussian mixture model clustering algorithm.

Top-down generation of low-resolution representations improves visual perception and imagination

The paper shows that using low-resolution top-down signals improves visual perception and imagination, and enhances sketch generation in AI by training models on blurred images.

Tokenizing Foldable Protein Structures with Machine-Learned Artificial Amino-Acid Vocabulary

This paper introduces ProTokens, artificial amino-acid tokens that enable efficient compression and processing of protein structures for design and analysis.

TREVI: A Transcriptional Regulation-driven Variational Inference Model to Speculate Gene Expression Mechanism with Integration of Single-cell Multi-omics

TREVIXMBD integrates scRNA-seq and scATAC-seq data to model gene regulation, improving cell typing and batch correction while revealing TF-gene regulation variations.

ProtWave-VAE: Integrating autoregressive sampling with latent-based inference for data-driven protein design

This paper introduces ProtWave-VAE, a generative model combining VAE and autoregressive methods for protein design, showing it can infer functional patterns and generate synthetic proteins with experimental validation.

Diffusion in a quantized vector space generates non-idealized protein structures and predicts conformational distributions

flowVI: Flow Cytometry Variational Inference

flowVI is a deep generative model for integrating flow cytometry datasets, improving imputation and enabling atlas construction by modeling noise and batch effects.

Cell-free biosynthesis combined with deep learning accelerates de novo-development of antimicrobial peptides

Autoencoder design plus a roughly 24-hour cell-free screen tested 500 candidates over five rounds, finding 30 functional AMPs; six were broad-spectrum with no resistance emerging in E. coli.

IC-VAE: A Novel Deep Learning Framework for Interpreting Multiplexed Tissue Imaging Data.

The paper introduces IC-VAE, a deep learning framework that improves the analysis of multiplexed tissue imaging data by controlling shared information among latent subspaces, leading to better downstream analysis.