How can I use geometricus to conduct t-SNE dimensionality reduction for my own pdb files?

[xjhzjucas]

Hi developer, I am a beginner in protein structure study and I read your paper "A structural biology community assessment of AlphaFold2 applications". Thank you for your nice job!
I have a question about how to do shape-mer t-SNE dimensionality reduction for my own pdb files, which were developed using AF2. I want to use your method to assess the relationship between my protein structures rather than structures from PDB or AFDB.
Can you give me some suggestions? Thank you so much!

[Ninjani]

Hi, the code for this analysis is here: https://github.com/TurtleTools/alphafold-structural-space
Hope that helps!

[xjhzjucas]

Hi, the code for this analysis is here: https://github.com/TurtleTools/alphafold-structural-space Hope that helps!

Thank you for your help! Actually I have installed "alphafold-structural-space" on my server and I tried to run 3 ipynb files. Since I don't need to download pdb files from AFDB and PDB, all of my pdb files were developed by AF2 by myself, so I don't have UniProt annotations files and species information, I don't know how can I modify the code for just do shape-mer t-SNE dimensionality reduction for my own 5000 pdb files. Thank you so much!

[Ninjani]

You could modify the get_AF_shapemers function in src/make_data.py to read and operate on your files, and then could use the second notebook to run the topic modelling.