variational autoencoder

Generative AAV capsid diversification by latent interpolation

The paper uses a Variational Auto-encoder to generate diverse AAV capsid variants with potential immune evasion properties by interpolating in sequence space.

PepVAE3: variational autoencoder framework for antimicrobial peptide generation and activity prediction

PepVAE3 generates antimicrobial peptides using a variational autoencoder and predicts their activity based on sequence and MIC data, enabling controlled design of novel AMPs with experimental validation.

Efficient and scalable integration of single-cell data using domain-adversarial and variational approximation

DAVAE integrates single-cell data across samples, technologies, and modalities by removing batch effects and enabling transfer learning and cell type prediction.

Explanatory latent representation of heterogeneous spatial maps of task-fMRI in large-scale datasets

The paper learns interpretable latent representations of task-fMRI data from large-scale datasets, capturing demographic and behavioral variability through a three-dimensional autoencoder and normative modeling.

Parallel inference of hierarchical latent dynamics in two-photon calcium imaging of neuronal populations

The paper introduces VaLPACa, a variational ladder autoencoder that disentangles hierarchical dynamics in calcium imaging data by separating calcium flux from neural computation.

Expanding functional protein sequence space using generative adversarial networks

This paper introduces ProteinGAN, a GAN that generates functional protein sequences, showing 24% of generated sequences are soluble and active in vitro.

Generating functional protein variants with variational autoencoders

This paper uses variational autoencoders to generate functional protein variants of bacterial luciferase, showing high success rates in retaining activity despite significant sequence changes.

Low-dimensional learned feature spaces quantify individual and group differences in vocal repertoires

The paper uses a variational autoencoder to analyze vocal behavior in mice and zebra finches, showing it outperforms handpicked features in capturing variability and relationships.

Consistent cross-modal identification of cortical neurons with coupled autoencoders

The paper develops coupled autoencoders to identify cortical neurons across modalities and finds consistent cell types in transcriptomic and electrophysiological data.

RaptGen: A variational autoencoder with profile hidden Markov model for generative aptamer discovery

RaptGen generates aptamers using a variational autoencoder with a profile hidden Markov model decoder, showing effectiveness in creating novel sequences not present in training data.

Fast and precise single-cell data analysis using hierarchical autoencoder

This paper introduces a hierarchical autoencoder for scRNA-seq analysis, showing improved performance in cell segregation, visualization, classification, and pseudo-time inference compared to existing methods.

pmVAE: Learning Interpretable Single-Cell Representations with Pathway Modules

pmVAE learns interpretable single-cell representations by incorporating pathway structure, improving detection of perturbation-targeted pathways in single-cell RNA-seq data.

Fast Variational Alignment of non-flat 1D Displacements for Applications in Neuroimaging

The paper introduces a variational method for aligning 1D signals with non-flat displacements, achieving state-of-the-art alignment quality and speed in neuroimaging applications.

Variational autoencoders learn universal latent representations of metabolomics data

This study trains a variational autoencoder on metabolomics data to learn nonlinear latent representations and validates its generalizability across different disease cohorts.

Binning microbial genomes using deep learning

This paper introduces a deep learning method using variational autoencoders to improve microbial genome binning from metagenomic data, achieving higher precision and completeness in reconstructed bins.

Combining Attention-based Multiple Instance Learning and Gaussian Processes for CT Hemorrhage Detection

This paper proposes an attention-based multiple instance learning approach for CT hemorrhage detection using scan-level labels, achieving better performance than methods relying on slice-level annotations.

Visualizing Population Structure with Variational Autoencoders

This paper uses variational autoencoders to visualize population genetic structure, showing better global geometry preservation than t-SNE and UMAP.

Probabilistic Harmonization and Annotation of Single-cell Transcriptomics Data with Deep Generative Models

This paper introduces scVI and scANVI, deep generative models for integrating single-cell RNA-seq data and annotating cell states, showing improved accuracy and scalability over existing methods.

Molecular free energy optimization on a computational graph

This paper introduces a computational graph framework for free energy optimization in molecular systems, achieving efficient and accurate protein structure refinement.

Haplotype and Population Structure Inference using Neural Networks in Whole-Genome Sequencing Data

HaploNet uses neural networks to infer population structure from whole-genome sequencing data, outperforming standard methods in distinguishing closely related populations.

Biological network-inspired interpretable variational autoencoder

This paper introduces VEGA, a variational autoencoder with interpretable latent variables inspired by biological networks, demonstrating its ability to model cellular responses and regulatory mechanisms.

DeepCellState: an autoencoder-based framework for prediction of cell type specific transcriptional states induced by drug treatment

DeepCellState predicts cell type-specific transcriptional states after drug treatment using an autoencoder, showing improved accuracy over baselines and generalizing to new cell types and interventions.

Gene set inference from single-cell sequencing data using a hybrid of matrix factorization and variational autoencoders

The paper introduces a hybrid model combining matrix factorization and variational autoencoders to infer gene sets from scRNA-seq data, correcting batch effects and identifying cell types.

AMPGAN v2: Machine Learning Guided Discovery of Anti-Microbial Peptides

AMPGAN v2 generates novel antimicrobial peptides with higher precision and selectivity for resistant targets using a bidirectional conditional generative adversarial network.

Learning to Encode Cellular Responses to Systematic Perturbations with Deep Generative Models

This study uses deep generative models to learn representations of transcriptomic data from cells under various perturbations, revealing relationships between perturbagens and drug-target mappings.

Design of an Epitope-Based Peptide Vaccine against the Severe Acute Respiratory Syndrome Coronavirus-2 (SARS-CoV-2): A Vaccine-informatics Approach

This study identifies 15 antigenic peptides from SARS-CoV-2 surface glycoprotein using vaccine-informatics, with 12 showing high similarity to SARS-CoV epitopes and potential for T-cell activation.

Latent Evolutionary Signatures: A General Framework for Analyzing Music and Cultural Evolution

This paper introduces a framework to identify evolutionary processes in music by extracting latent signatures, showing improved performance in period and genre prediction.

Learning to leverage salient regions in neuro-oncology using Deep Learning

This paper compares two deep learning methods for localizing brain tumors in MRI images using BraTS and internal datasets, showing both can localize tumors to some extent.

De Novo Protein Design for Novel Folds using Guided Conditional Wasserstein Generative Adversarial Networks (gcWGAN)

This paper develops a deep generative model, gcWGAN, to design novel protein folds using sequence-structure data, achieving higher fold coverage and sequence diversity than a competing method.

NuSeT: A Deep Learning Tool for Reliably Separating and Analyzing Crowded Cells

NuSeT improves nuclear segmentation in microscopy images by combining U-Net and RPN with watershed, reducing false positives and handling complex overlaps.