search

rdk / p2rank

P2Rank: Protein-ligand binding site prediction tool based on machine learning. Stand-alone command line program / Java library for predicting ligand binding pockets from protein structure.
https://rdk.github.io/p2rank/
MIT License
242 stars 34 forks source link

Fpocket in tutorial #25

Closed skodapetr closed 4 years ago

skodapetr commented 4 years ago

Is it necessary to mention historical reason and Fpocket in the training-tutorial.md ? I though p2rank provide better results and as a user, this information is not necessary for me making the tutorial harder to read.

rdk commented 4 years ago

I still do consider it important, at least until the change of P2Ranks default model and associated parameters. Otherwise you might wonder why training P2Rank with default parameters will not give as good results as presented in papers.

This is key:

I recent versions it might be possible to achieve better results by training from whole protein surface in combination with class balancing techniques (see the next section). Note that default values of other parameters (related to feature extraction and classification results aggregation) were optimized for the case where sample_negatives_from_decoys = true. ... Their values may need to be optimized again for case of sample_negatives_from_decoys = false.