cellppd

cellppd predicts and designs cell penetrating peptides (CPPs) to enable computational screening and discrimination of CPPs for intracellular delivery of cargoes such as oligonucleotides, small molecules, and proteins.

Key Features:

• SVM models: Support vector machine (SVM)-based predictive models trained on a dataset of 708 experimentally validated CPPs.
• Input features: Uses amino acid composition, dipeptide composition, binary profile of patterns, and physicochemical properties as model inputs.
• Motif identification: Identification of specific motifs within CPPs incorporated into prediction.
• Hybrid prediction model: Combines motif information with binary profile features for improved prediction.
• Cross-validation performance: Achieves a maximum accuracy of 97.40% based on five-fold cross-validation.
• Independent evaluation: Demonstrates a maximum accuracy of 81.31% and a Matthew's Correlation Coefficient (MCC) of 0.63 on an independent dataset.
• Residue preferences: Reports positional preferences for Arg, Lys, Pro, Trp, Leu, and Ala that aid discrimination between CPPs and non-CPPs.

Scientific Applications:

• CPP prediction for intracellular delivery: Computationally predicts CPPs suitable for delivering oligonucleotides, small molecules, and proteins into cells.
• Peptide design and optimization: Guides design and in silico optimization of candidate CPPs prior to experimental synthesis.
• Classification of peptides: Discriminates cell-penetrating from non-cell-penetrating peptides based on composition and motif features.

Methodology:

SVM-based models trained on 708 experimentally validated CPPs using amino acid composition, dipeptide composition, binary profile of patterns, and physicochemical properties; specific motifs were identified and integrated into a hybrid model combining motif information with binary profiles, with performance assessed by five-fold cross-validation and independent dataset evaluation.