NLStradamus

NLStradamus predicts nuclear localization signals (NLSs) in proteins using hidden Markov models to identify sequence patterns that mediate regulated nuclear import.

Key Features:

• Hidden Markov model prediction: Uses hidden Markov models (HMMs) to model and identify NLS sequence motifs.
• NLS detection: Predicts nuclear localization signals within protein sequences, including motifs associated with the classical NLS pathway that involve basic residue patterns.
• Yeast-derived patterns: Built from an analysis of characterized NLSs in yeast, revealing consistent amino acid residue patterns across import pathways.
• Novel NLS identification: Identifies novel NLSs by modeling conserved residue patterns across known examples.
• Performance metrics: Consistently detects 37% of characterized NLSs while maintaining a low false positive rate.
• Robustness: Maintains predictive performance when applied to datasets outside those used for training.
• Addresses prior limitations: Targets the low true positive rates reported for existing computational NLS prediction methods by leveraging HMM-based sequence modeling.

Scientific Applications:

• Protein localization studies: Aids identification of proteins targeted to the nucleus by predicting candidate NLS motifs.
• Nuclear transport research: Supports investigation of nuclear import mechanisms, including the classical import pathway mediated by basic-residue NLSs.
• Cross-species NLS discovery: Can be applied to datasets from organisms beyond yeast to propose putative NLS-containing proteins.

Methodology:

Computational methods include analysis of characterized NLSs in yeast and HMM-based sequence modeling to identify NLS motifs, with evaluation reporting a 37% detection rate and a low false positive rate on external datasets.