IL-6Pred

IL-6Pred predicts interleukin-6 (IL-6) inducing peptides to enable identification and design of peptides that modulate IL-6-mediated immune responses.

Key Features:

• Prediction Model: Models were trained on a dataset of 365 experimentally validated IL-6 inducing peptides and 2,991 non-inducing peptides sourced from the Immune Epitope Database.
• Feature Computation: Each peptide was characterized by 9,149 features computed using Pfeature.
• Feature Selection and Reduction: Features were reduced from 9,149 to 186 using the SVC-L1 technique, and the top 10 features were selected based on classification ability.
• Machine Learning Performance: Multiple algorithms were evaluated and a Random Forest model achieved AUROC values of 0.84 on the training dataset and 0.83 on an independent validation dataset.
• Application to SARS-CoV-2: The approach was applied to identify IL-6 inducing peptides within SARS-CoV-2 proteins for use in studies related to COVID-19 immune responses.

Scientific Applications:

• Vaccine and epitope design: Identification of IL-6 inducing peptides can inform selection of epitopes for vaccine design and immunogen engineering.
• Immunology research: Predicted IL-6 inducing peptides support studies of IL-6-driven inflammatory and immune response mechanisms.
• Disease-specific investigation: Predictions can be used to explore IL-6 involvement in COVID-19 and other inflammatory conditions characterized by elevated IL-6.

Methodology:

Models were trained on 365 IL-6 inducing and 2,991 non-inducing peptides from the Immune Epitope Database; 9,149 features per peptide were computed with Pfeature, reduced to 186 via SVC-L1, the top 10 features were selected for classification, and multiple machine learning algorithms including Random Forest were evaluated (Random Forest AUROC 0.84 training, 0.83 independent validation).