pLoc_bal-mEuk

pLoc_bal-mEuk predicts the subcellular localization of eukaryotic proteins from sequence information to provide accurate multi-label localization assignments while mitigating biases from imbalanced training datasets.

Key Features:

• Subcellular localization prediction: Predicts localization of eukaryotic proteins using sequence-derived information.
• Multi-label assignment: Supports multi-label systems for proteins that localize to multiple subcellular compartments.
• General PseAAC encoding: Represents protein sequences using General PseAAC (Pseudo Amino Acid Composition).
• Quasi-balanced training dataset: Employs a quasi-balancing approach to reduce bias from skewed class distributions, addressing imbalances reported up to 200-fold.
• Cross-validation evaluation: Uses cross-validation tests to evaluate and confirm improved performance relative to pLoc-mEuk.
• Improved accuracy and reliability: Enhances prediction accuracy and reliability through quasi-balancing of training data.

Scientific Applications:

• Protein localization studies: Supports research in protein subcellular localization within eukaryotic systems.
• Post-genomic functional annotation: Facilitates functional annotation of proteins in post-genomic analyses with expanding sequence datasets.
• Drug development research: Provides localization information useful for research and analyses in drug development.
• Datasets with class imbalance: Applicable to biological systems and datasets that exhibit severe class imbalances.

Methodology:

Uses General PseAAC (Pseudo Amino Acid Composition) to encode sequences and trains on a quasi-balanced training dataset to address biases in skewed data distributions, with performance assessed by cross-validation.