search

samuelmurail / af2_analysis

Analysis of alphafold and colabfold results
GNU General Public License v2.0
8 stars 2 forks source link

Confidence on LIS scoring ? #4

Open tubiana opened 3 weeks ago

tubiana commented 3 weeks ago

Hi @samuelmurail :)

Just asking if you found any inconsistencies in the computation of the LIS score ?

Here's for example the LIS score matrix from a model I made with af2_analysis : image

And the LIS score from the original notebook : image

Or maybe there is something I misunderstood on af2_analysis LIS matrix :) ?

Cheers, Thibault

samuelmurail commented 2 weeks ago

Hi @tubiana

I tested it on my side. I guess you are not using the same versions of the library af2_analysis in the two cases. I made a lot of change before summer holidays, that may explain your issue.

Here is how to use it:

# Import your models
my_data = af2_analysis.Data('./src/af2_analysis/test/inputs/beta_amyloid_dimer_cf_1.5.5/', verbose=False)
#compute_LIS_pep
docking.LIS_pep(my_data, verbose=False)

You should have a matrix for the LIS field of size $N \times N$ where $N$ is the number of chain of your model. And you should also have access to the LIS_rec_pep LIS_pep_rec values which are floating values.

I have updated the example.ipynb if it can help. Let me know if your inconsistencies are fixed.

Cheers, Samuel