Pre- and Post-publication Verification for Reproducible Data Mining in Macromolecular Crystallography

Area of Science:

• Structural biology
• Biophysics
• Computational biology

Background:

• Scientific publications rely on peer review for validation, but the underlying data also requires rigorous scrutiny.
• Reproducibility of studies and their conclusions is impossible without access to and verification of the underpinning data.
• The Protein Data Bank (PDB) provides open access to structural data, facilitating numerous applications.

Purpose of the Study:

• To emphasize the critical need for data verification alongside narrative review in scientific publications.
• To highlight the role of reproducible data in scientific evaluation and replication.
• To discuss the integration of artificial intelligence and machine learning in scientific validation.

Main Methods:

• Review of current practices in scientific data handling and publication.
• Discussion of data generation methods, including X-ray crystallography, X-ray free electron lasers (XFELs), electron cryomicroscopy (cryoEM), and neutron diffraction.
• Exploration of the potential of artificial intelligence (AI) and machine learning (ML) in scientific validation.

Main Results:

• The accuracy of structural determination methods is confirmed through reproducibility by independent methods.
• Increasing data generation rates at advanced facilities necessitate robust validation processes.
• Examples involving rhenium theranostics, anti-cancer platins, and SARS-CoV-2 main protease illustrate the importance of data integrity.

Conclusions:

• Scrutiny of underpinning data is as crucial as narrative review for establishing a version of record in scientific publications.
• Reproducible data is the foundation for reliable data mining and scientific replication.
• Future scientific validation may involve AI and ML complementing human expertise for enhanced pre-publication review.