StackCPPred

StackCPPred predicts cell-penetrating peptides (CPPs) and estimates their cellular uptake efficiency to support selection of CPPs for delivery of pharmacologically active molecules such as short interfering RNAs, nanoparticles, plasmid DNAs, and small peptides.

Key Features:

• RECM-composition: Encodes residue composition based on the pairwise energy content of residues.
• PseRECM: Extends RECM-composition by incorporating sequence-order information derived from pairwise residue energies.
• RECM-DWT: Applies discrete wavelet transform to RECM-derived signals to capture complex patterns in peptide sequences.
• Stacking-based machine learning: Integrates RECM-composition, PseRECM, and RECM-DWT features within a stacking ensemble to predict CPPs and uptake efficiency.
• Performance metrics: Reports accuracies of 94.5% on the CPP924 dataset and 78.3% on the CPPsite3 dataset, outperforming comparative predictors by 2.9% and 5.8%, respectively.
• Validation: Performance assessed using jackknife validation on the CPP924 and CPPsite3 datasets.

Scientific Applications:

• Facilitating experimental design: Provides predictions of CPP presence and uptake efficiency to guide peptide annotation and selection for experimental testing.
• Supporting therapeutic delivery research: Informs selection of CPPs for delivery of short interfering RNAs, nanoparticles, plasmid DNAs, and small peptides in preclinical studies.

Methodology:

Pairwise energy content of residues is encoded via RECM-composition, PseRECM, and RECM-DWT features, which are integrated into a stacking-based machine learning model and evaluated by jackknife validation on CPP924 and CPPsite3.