gcWGAN

The software tool 'gcWGAN' employs semisupervised guided, conditional, Wasserstein Generative Adversarial Networks to predict protein sequences for novel structural folds. Using Wasserstein distance in the loss function, it builds on conditional Wasserstein GAN to overcome training challenges. Key features include a low-dimensional fold space representation, an ultrafast sequence-to-fold predictor, and semisupervised training using sequence data.

Topic

Protein folding, stability and design;Protein folds and structural domains;Physics;Structure prediction

Detail

  • Operation: Fold recognition;Protein folding analysis;Ab initio structure prediction

  • Software interface: Command-line user interface

  • Language: Python

  • License: The GNU General Public License v3.0

  • Cost: Free

  • Version name: -

  • Credit: the National Institutes of Health (NIH).

  • Input: -

  • Output: -

  • Contact: Yang Shen yshen@tamu.edu, Mostafa Karimi mostafa_karimi@tamu.edu, Shaowen Zhu shaowen1994@tamu.edu, Yue Cao cyppsp@tamu.edu

  • Collection: -

  • Maturity: -

Publications

  • De Novo Protein Design for Novel Folds Using Guided Conditional Wasserstein Generative Adversarial Networks.
  • Karimi M, et al. De Novo Protein Design for Novel Folds Using Guided Conditional Wasserstein Generative Adversarial Networks. De Novo Protein Design for Novel Folds Using Guided Conditional Wasserstein Generative Adversarial Networks. 2020; 60:5667-5681. doi: 10.1021/acs.jcim.0c00593
  • https://doi.org/10.1021/ACS.JCIM.0C00593
  • PMID: 32945673
  • PMC: PMC7775287

Download and documentation

< Back to DB search