IDEPI

IDEPI predicts HIV-1 phenotypic traits from viral sequence data to identify genotype-to-phenotype relationships relevant to antibody neutralization, coreceptor tropism, compartmentalization, and drug-resistance mutations.

Key Features:

• Machine Learning Platform: Uses open-source general-purpose machine learning algorithms and libraries to build predictive models correlating genotypes with phenotypes.
• Phenotype Prediction: Applies trained models to classify sequences with unknown phenotypes and predict traits such as susceptibility to neutralization by specific antibodies.
• Feature Identification: Identifies specific sequence features that contribute to phenotypic outcomes, enabling interpretation of genetic determinants.
• Implementation: Provided as cross-platform Python source code for computational use.

Scientific Applications:

• Epitope Prediction for Broadly Neutralizing Antibodies (bNab): Identifies epitopes recognized by broadly neutralizing antibodies to inform vaccine-related studies.
• Coreceptor Tropism Determination: Predicts HIV-1 coreceptor usage to inform studies of viral entry and pathogenesis.
• Compartment-Specific Genetic Signatures: Detects genetic signatures specific to host compartments to study viral compartmentalization.
• Drug-Resistance Mutation Deduction: Deduces mutations associated with drug resistance to inform treatment-relevant analyses.

Methodology:

IDEPI trains machine learning models on sequence data with known phenotypes to learn genotype-to-phenotype relationships and then applies these models to new sequences to predict phenotypes and identify contributing genetic elements.