Plant-mSubP

Plant-mSubP predicts single- and dual-target subcellular localizations for plant proteins to support annotation of plant proteomes and studies of cellular functions.

Key Features:

• Prediction scope: Predicts 11 single subcellular localizations and three dual localizations for plant proteins.
• Feature representation: Represents proteins using hybrid features based on composition and physicochemical properties.
• Amino acid composition: Uses amino acid composition as an input feature.
• Pseudo amino acid composition (PseAAC): Employs PseAAC to capture sequence-order information and physicochemical properties.
• Auto-correlation descriptors: Incorporates auto-correlation descriptors to model residue property correlations along sequences.
• Quasi-sequence-order descriptors: Uses quasi-sequence-order descriptors to encode sequence-order effects.
• PseAAC-NCC-DIPEP hybrid features: Utilizes hybrid features such as N-Center-C terminal amino acid composition combined with dipeptide composition (PseAAC-NCC-DIPEP).

Scientific Applications:

• Plant proteome annotation: Supports annotation of plant proteomes by predicting single and multiple subcellular localizations.
• Cellular function inference: Enables investigation of protein localization to inform studies of cellular functions and processes in plants.

Methodology:

The method uses hybrid feature sets (composition and physicochemical descriptors including PseAAC, auto-correlation, quasi-sequence-order, and PseAAC-NCC-DIPEP) to represent proteins and evaluates predictive performance on training and independent test datasets; training accuracies: 81.97% (single-label), 84.75% (combined single- and dual-label), 87.88% (dual-label); independent test accuracies: 64.36% (single-label), 64.84% (combined single- and dual-label), 81.08% (dual-label).