Skip to content

New benchmark: protein sampling #726

Description

@lwalew

Summary

This benchmark evaluates the quality and accuracy of Machine Learning Interatomic Potentials (MLIP) by analyzing the conformational sampling of amino acids in small proteins during molecular dynamics simulations. Specifically, it computes backbone Ramachandran angles (phi/psi) and side chain rotamer angles (chi1, chi2, …). The sampled probability distribution of these angles is then compared against reference data and outliers are detected.

Interactive features

No response

Category

No response

Data availability

No response

Computational cost

No response

Additional context

No response

Metadata

Metadata

Assignees

No one assigned

    Labels

    externalContribution from external frameworknew benchmarkProposals and suggestions for new benchmarks

    Type

    No type

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions