iNR-PhysChem

iNR-PhysChem predicts whether a protein sequence is a nuclear receptor (NR) and assigns identified NRs to specific NR subfamilies for sequence-based functional classification.

Key Features:

• Two-Level Prediction: Operates in two stages to detect NRs and then classify identified NRs into subfamilies.
• First Level (NR identification): Determines whether an input protein sequence is a nuclear receptor (NR).
• Second Level (Subfamily classification): Classifies identified NRs into one of seven subfamilies: Thyroid hormone like (NR1), HNF4-like (NR2), Estrogen like, Nerve growth factor IB-like (NR4), Fushi tarazu-F1 like (NR5), Germ cell nuclear factor like (NR6), and Knirps like (NR0).
• Fuzzy K Nearest Neighbor (FK-NN) classifier: Uses an FK-NN classifier informed by pseudo amino acid composition.
• Feature types: Employs amino acid composition, dipeptide composition, complexity factor, and low-frequency Fourier spectrum components as input features.
• Performance: Reported accuracies on benchmark datasets are approximately 93% for NR identification (first level) and 89% for subfamily classification (second level).

Scientific Applications:

• NR identification from sequence data: Enables identification of nuclear receptors among uncharacterized protein sequences.
• NR subfamily assignment: Provides subfamily-level classification of identified NRs for sequence-based annotation.

Methodology:

Two-level classification using a Fuzzy K Nearest Neighbor (FK-NN) classifier with pseudo amino acid composition features derived from amino acid composition, dipeptide composition, complexity factor, and low-frequency Fourier spectrum components.