MichiGAN: sampling from disentangled representations of single-cell data using generative adversarial networks Long Description (required)
MichiGAN is a novel neural network that combines the strengths of VAEs and GANs to sample from disentangled representations without sacrificing data generation quality. MichiGAN allows us to manipulate semantically distinct aspects of cellular identity and predict single-cell gene expression response to drug treatment. Cellular identity, Disentangled representations, Generative adversarial networks, Representation learning, Single-cell genomics https://github.com/welch-lab/MichiGAN
Step 1: Use the attribute tree to add new attributes or remove pre-selected attributes to describe the simulator.
Every sub-attribute is selected Not all sub-attributes are selectedFill Clear Expand Collapse Reset
Summary of Proposed Changes Step 2: Review list of proposed attribute addition(s) and subtraction(s).
Can't Find the Attribute You Are Looking For? If you would like to propose an attribute that you cannot find in the tree above, or if you would like to add a clarification to one or more attributes for this simulator (e.g. a specific file format for attribute /Output/File Format/Other), please list them in the Additional Comment box of the Submit tab .
Summary of Proposed Changes Current Citations/Applications
[Pubmed ID: 34016135 ],
Yu H, Welch JD ,
MichiGAN: sampling from disentangled representations of single-cell data using generative adversarial networks. ,
Genome Biol ,
05-20-2021 ,
https://www.ncbi.nlm.nih.gov/pubmed/?term=34016135, Primary Citation
[Pubmed ID: 35409058 ],
Hazra D, Kim MR, Byun YC ,
Generative Adversarial Networks for Creating Synthetic Nucleic Acid Sequences of Cat Genome. ,
Int J Mol Sci ,
03-28-2022 ,
https://www.ncbi.nlm.nih.gov/pubmed/?term=35409058, , Application