IL-13Pred

IL-13Pred predicts Interleukin-13 (IL-13) inducing peptides to identify peptide sequences that modulate IL-13–mediated immune responses implicated in airway hyperresponsiveness, glycoprotein hypersecretion, goblet cell hyperplasia, inhibition of tumor immunosurveillance, carcinogenesis, and severe COVID-19.

Key Features:

• Experimental dataset: Uses 313 experimentally validated IL-13 inducing peptides and 2,908 non-inducing Homo sapiens peptides sourced from the Immune Epitope Database (IEDB).
• Feature generation: Generates 9,165 sequence-derived features using Pfeature.
• Feature selection: Applies SVC-L1 to extract 95 key features and ranks them for model building, with the top 10 features used for final models.
• Machine learning models: Implements various machine-learning techniques, with XGBoost reported as the best-performing algorithm.
• Performance metrics: The XGBoost model achieved maximum AUCs of 0.83 on the training dataset and 0.80 on an independent dataset.
• Model validation: Employs five-fold cross-validation for training, testing, and evaluation to assess model robustness.
• Application to viral variants: Analysis indicates certain SARS-CoV-2 variants may have higher propensity to induce IL-13, informing peptide-level assessments.

Scientific Applications:

• Therapeutic design: Identification of IL-13 inducing peptides to guide design and safety assessment of protein therapeutics and immunotherapies.
• Disease mechanism studies: Characterization of peptides that influence IL-13 activity to study molecular mechanisms in allergic diseases and cancer.
• Vaccine candidate screening: Screening of vaccine antigen sequences for potential IL-13 induction to inform vaccine safety evaluation, including SARS-CoV-2 analyses.

Methodology:

Compiled positive (313) and negative (2,908) peptide datasets from IEDB, generated 9,165 features with Pfeature, applied SVC-L1 to select 95 features and ranked them to use the top 10 for machine-learning models including XGBoost, and evaluated models using five-fold cross-validation and an independent dataset.