APPTEST

APPTEST predicts peptide tertiary structures from primary amino acid sequences using neural network architectures and simulated annealing, enabling accurate modeling of linear and cyclic peptides (5–40 natural amino acids).

Key Features:

• Neural network architectures: Employs neural network-based models to derive structural information from primary amino acid sequences.
• Simulated annealing: Applies simulated annealing methods to generate tertiary structure conformations.
• Sequence input: Accepts primary amino acid sequences as the input for prediction.
• Peptide scope: Handles both linear and cyclic peptides composed of 5–40 natural amino acids.
• Prediction speed: Produces predicted structures within minutes.
• Benchmark evaluation: Evaluated on a test dataset of 356 peptides.
• Performance metrics: Achieved an average backbone deviation of 1.9Å and identified native or near-native structures in 97% of targets.
• Comparative benchmarking: Produced more native-like structures than PEP-FOLD, PEPstrMOD, and Peplook across short, long, and cyclic peptide benchmark datasets.

Scientific Applications:

• Peptide tertiary structure prediction: Modeling tertiary structures of linear and cyclic peptides for structural biology studies.
• Peptide design: Informing in silico peptide design through predicted structural models.
• Functional and interaction analysis: Supporting analysis of peptide function and interactions with biological targets.

Methodology:

Generates predictions from primary amino acid sequences using neural network architectures and simulated annealing and was evaluated on a test set of 356 peptides with reported average backbone deviation of 1.9Å and 97% native/near-native identification.