pepp'r

I have a structure prediction model and now I want to know how well it performs in reproducing the reference structures. But there are so many possible metrics, some for monomers, some for complexes! Is there a package that handles this for me?

Try

It's a Package for Evaluation of Predicted Poses, Right?

Yes, indeed! It allows you to compute a variety of metrics on your structure predictions for assessing their quality. It supports

• all CASP/CAPRI metrics and more
• small molecules to huge protein complexes
• easy extension with custom metrics
• a command line interface and a Python API

Installation

peppr is available via PyPI:

$ pip install peppr

Usage example

Using the CLI, you can either compute a single metric for a system...

$ peppr run dockq reference.cif poses.cif

... or run an entire prediction model evaluation on many systems.

# Select the metrics you want to compute (here: RMSD and lDDT)
$ peppr create peppr.pkl monomer-rmsd monomer-lddt

# Run the evaluation on predicted poses and their corresponding references
$ peppr evaluate-batch peppr.pkl "systems/*/reference.cif" "systems/*/poses"

# Select the aggregation method over poses (here: Top-3 and Oracle) and report the results
$ peppr tabulate peppr.pkl table.csv top3 oracle

Available metrics

• RMSD
• TM-score
• lDDT
• lDDT-PLI
• fnat
• iRMSD
• LRMSD
• DockQ

... and more!