Downloading Data

[collinarnett]

Figuring out how to download data is the first issue faced when trying to implement this paper. The original authors noted in section 3.1 that the dataset was obtained from Protein Data Bank and they also noted that the data downloaded was the 3D structure of the protein so that leads me to believe that the authors used the download option on the PDB website.

Although they haven't specified which download option they decided on we can infer that they used the Coordinates & Experimental Data option with the Structural Factors box ticked from their wording:

We chose to encode 3Dstructure as 2D pairwise distances between↵-carbons on the protein backbone. This representation does not preserve information about the protein sequence (side chains) or the torsion angles of thepolypeptide backbone, but preserves enough information to allow for structure recover

However they do include the train and test data in their supplementary files listed here

[collinarnett]

The supplemental files are as follows:

model_architectures.txt
supplement.pdf
test_ids.txt
train_ids.txt

Now we can focus on automating the download process using the provided ids before figuring out the model architecture or the supplement pdf.

[collinarnett]

here is a very helpful link in easily downloading these files programatically.

[collinarnett]

Finished downloading the data and uploaded the data to my personal AWS for easy downloads later. TODO: Looking for a place to publicly host the compressed train and test set.