search

phbradley / alphafold_finetune

Python code for fine-tuning AlphaFold to perform protein-peptide binding predictions
Apache License 2.0
139 stars 20 forks source link

Binding Score #14

Open mesdaghi opened 6 months ago

mesdaghi commented 6 months ago

Hi, great work. I've managed to succefully run a prediction and obtained the following confidence scores:

model_2_ptm_ft_plddt model_2_ptm_ft_pae model_2_ptm_ft_plddt_0 model_2_ptm_ft_pae_0_0 model_2_ptm_ft_pae_0_1 model_2_ptm_ft_plddt_1 model_2_ptm_ft_pae_1_0 model_2_ptm_ft_pae_1_1

How are these converted to the 'binding score' that is described in the paper? Thank you.

phbradley commented 6 months ago

Hi there, the MHC-peptide PAE can be computed from the .npy file that Alphafold saves with all the residue-residue PAE scores. Or you can take the average of model_2_ptm_ft_pae_0_1 (ie, the chain 0 to chain 1 PAE) and model_2_ptm_ft_pae_1_0 (ie, the chain 1 to chain 0 PAE):

0.5 * (model_2_ptm_ft_pae_0_1 + model_2_ptm_ft_pae_1_0)