colab_embedding.md

Distance-AF embedding generation

1. Google colab(suggested)

Running instructions

1. Go to Distance_AF_embedding.ipynb.
2. At the right top, click 'Open in Colab'.
3. Install the environment and dependency by running the first cell '1. Install third-party software' and second cell '2. Download AlphaFold'.
4. Prepare your sequence and job name at '3. Enter the amino acid sequence(s) to fold'.
   • REQUIRED: filled out your sequence at the text box sequence_1.
   • OPTIONAL: change the job_name text box with any name you like.
   • OTHERS: keep empty for sequence 2, sequence 3, etc for current version.
5. Run the thrid cell '3. Enter the amino acid sequence(s) to fold', fourth cell '4. Search against genetic databases' and fifth cell '5. Run AlphaFold and download embeddings' consecutively.
6. The embedding zip file will be downloaded automatically named in 'job_name.zip'

More details

7. In the zip file, it contains model_1.npz, model_2.npz, model_3.npz, model_4.npz, model_5.npz, corresponding to the embedding file derived from 5 models from AlphaFold. The PDB files model_1.pdb, model_2.pdb, model_3.pdb, model_4.pdb, model_5.pdb are the AlphaFold2 predicted structure.
8. To choose more appropriate embedding file among the 5 npz files, you can compare your initial structure with the 5 PDB files, choose the corresponding npz file which is closest to your initial structure. For example, you find model_2.pdb is the closest one, you then choose model_2.npz as the embedding file.

2. Run Embedding generation locally(on going...)