- Explore protein encoding (https://github.com/ethanmoyer/ptnstrerrpredict/blob/master/ptn.py)
- Data set is through Protein Data Bank through API calls (https://www.wwpdb.org/)
Paper 1: Functional Protein Structure Annotation using a Deep Colvolution Generative Adversarial Network
- Create a deep convolution generalized advesarial network (DCGAN) to generate and discriminate between functional and non-functional protein structures.
Conference: https://www.bhi-bsn-2021.org/
Paper 2: Novel Protein Structure Generation according to Specific Function using a Deep Colvolution Generative Adversarial Network
- Use a DCGAN to generate and discriminate between proteins with a particular function, such as ligand binding, RNA cleaving, etc
- Iteratively train and test an autoencoder network on increadinly messed-up decoy protein structures and explore the extent to which an autoencoder can refold them.
[x] - Define protein structures of interest, search space, and data base focus ()
[x] - Build database using list of four-letter protein structure IDs ()
[] - Download proteins from PDB and encode protein structures into grid structure (Ethan, )
[x] - Determine feature set based on encoding and decide if any more features should be included ()
[] - Train GAN/DCGAN on encoded protein structures ()
[] - Perform experiemnts/gather results to determine how well the model performed ()
[] - Explore the explainability of the model ()
[] - Generate figures for paper and sample test cases to illistrate how the model performs on an individual protein structure ()
Ethan: Data set, protein encoding, Isamu: Results Jeff: background/GAN architecture, torsion angle Mali: Introduction Alisha: Background of protein structure prediction, abstract Yigit: Related work/GAN architecture Adam: Background of protein structure prediction
- Introduce GANs (Jeff)
- Introduce protein structure encoding, prediction, mapping (Ethan)
- Measuring Protein Structure using Machine Learning:
- https://www.mabc2020.com/post/measuring-protein-structure-deviation-using-machine-learning
- Introduce GitHub for project management (Yigit)
- Assign taks for paper based on individual interest (Ethan)
- Aim to have paper #1 on DCGAN for functional vs nonfunctional protein structure annotation done and submitted by Sunday (18 April 2021)
- Discuss autoencoder project and more functional annotation of protein