NucleoSeeker

NucleoSeeker curates high-quality, tailored RNA datasets from the Protein Data Bank (PDB) by applying multiple filters at structure, sequence, and annotation levels to support training and evaluation of RNA structure prediction models.

Key Features:

• Structure-level filtering: Filters RNA entries based on structural criteria to ensure data quality.
• Sequence-level filtering: Filters RNA entries based on sequence criteria to control dataset composition.
• Annotation-level filtering: Filters RNA entries based on annotation criteria to refine dataset selection.
• Nonredundant dataset creation: Enables generation of nonredundant RNA structure datasets for benchmarking and model assessment.
• Unified framework: Combines multiple tools into a single pipeline for streamlined data curation.

Scientific Applications:

• RNA structure prediction model training: Creates tailored, high-quality datasets for training novel deep learning models for RNA structure prediction.
• RNA structure prediction model evaluation: Curates nonredundant datasets to assess performance of existing methods such as AlphaFold3.
• Benchmarking: Facilitates creation of standardized datasets to compare and judge RNA structure prediction methods.

Methodology:

NucleoSeeker combines multiple tools into a unified framework that applies filters at structure, sequence, and annotation levels to curate datasets from the Protein Data Bank (PDB).