html1101 / Science-Fair-2020-2021

mRNA Sequence Design Using Optimization Techniques
1 stars 2 forks source link

How to use this project? #3

Open PlantBiTree opened 2 years ago

PlantBiTree commented 2 years ago

Dear engineer: I downloaded your project <Science Fair 2020-2021>(which is about codon optimization).Then I tried to run the project in Pycharm. But I didn't figure out how to make a graphic like CASP_TEST.png and rnaStructureVaccine.png. So could you tell me how to use this project and how to make these graphis? By the way I want to realize how to run the web service.I would appreciate that if you could tell me how to operate them.My email add is bbzyp88@gmail.com. Thank you soooooooo much!

Looking forward your reply.

html1101 commented 2 years ago

Hi @PlantBiTree, thanks for taking a look at my project!

Apologies in advance if my response isn't super helpful, it's been a year so I've forgot a little bit about the structuring of the repo :). At the time the repo was really just a total mind splurge, so I've just committed and removed some random files I had placed in there that I didn't use in the final project.

Basically, the brunt of the work is done in the file discrete_optimization.py. The idea is that you can take the antigen of essentially any virus (I looked at the SARS-CoV-2 spike protein, since it was pretty relevant at the time), get its sequence, and then run a series of optimizations such that you can create an mRNA vaccine, which is the mRNA antigen sequence, but with some key adjustments that allows your body to better recognize hey, I need to make antibodies for that. I added a couple of existing methods to solve this problem (codon_mapping.py and codon_table_sol.py), which are a lot faster, but using discrete optimization warrants better results (90.7% of the discrete optimization sequence is identical to the Pfizer vaccine).

The images are largely visualizations I used in the research paper. CASP_TEST.png is a visualization of a protein structure simulation software, AlphaFold, I describe in my background research; basically, I just got the original structure of the protein 5chy, then predicted it using AlphaFold and compared them visually using VMD. My initial project was to find the structure of antibodies that fit a particular antigen, but, because it was too computationally expensive, I didn't end up using molecular graphics simulators in the final version, and finding the antibodies that correspond to a particular antigen isn't as effective as using mRNA vaccines.

The "application_viewer" is part of the old antibody version of the project, but I'll keep it there if you find some value in it; basically, you just need to open up main.html in a browser window, then you can select a pdb file and it'll render the protein. If you wanted to visualize a protein, though, I'd recommend just using the built-in browser protein viewer available here.