Dear All,
You may have seen our announcement today about expanding the AlphaFold Protein
Structure Database to 214M predicted models. To enable this expansion, we’ve
updated the Predicted Aligned Error (PAE) JSON format to make it compact (about
4x smaller):
The PAE JSON numbers are now rounded to the closest integer, giving ~75%
compressed size reduction. The integer resolution is sufficient for analytical
purposes.
The indices are not stored anymore since we store the full 2D PAE matrix rather
than a sparse one, giving ~4% compressed size reduction.
The “distances” field has been renamed to “predicted_aligned_error” and is now
stored as a 2D array of shape (num_res, num_res) rather than a 1D array. We
renamed the field on purpose so that existing code breaks rather than
potentially silently returning wrong values.
For a protein of length num_res, the PAE JSON file has now the following format:
[{
"predicted_aligned_error": [[0, 1, 4, 7, 9, ...], ...], # Shape: (num_res,
num_res).
"max_predicted_aligned_error": 31.75 # Scalar.
}]
The fields in the JSON file are:
predicted_aligned_error: The PAE value of the residue pair, rounded to the
closest integer. For PAE value on position (i, j), i is the residue on which
the structure is aligned for the predicted error, j is the residue on which the
error is predicted.
max_predicted_aligned_error: A number that denotes the largest possible
unrounded value of PAE that could occur in the PAE array. The smallest possible
value of PAE is 0.
The updated PAE format is only available from the AlphaFold Protein Structure
Database. The PAE format from the AlphaFold Colab notebook is not updated.
If you require support with this change, please email [email protected]
<mailto:[email protected]> and they may be able to assist.
Best Wishes,
Sameer Velankar
########################################################################
To unsubscribe from the CCP4BB list, click the following link:
https://www.jiscmail.ac.uk/cgi-bin/WA-JISC.exe?SUBED1=CCP4BB&A=1
This message was issued to members of www.jiscmail.ac.uk/CCP4BB, a mailing list
hosted by www.jiscmail.ac.uk, terms & conditions are available at
https://www.jiscmail.ac.uk/policyandsecurity/
