SCOTCH: subtype A coreceptor tropism classification in HIV-1

Area of Science:

• Virology
• Computational Biology
• Infectious Diseases

Background:

• The V3 loop of Human Immunodeficiency Virus 1 (HIV-1) gp120 glycoprotein determines viral coreceptor tropism, binding either CCR5 or CXCR4.
• CCR5 is a target for entry inhibitors, making accurate tropism prediction crucial for effective antiretroviral therapy.
• Existing tropism prediction models primarily use subtype B sequences, showing reduced reliability for non-B subtypes, especially subtype A.

Purpose of the Study:

• To develop a reliable prediction model for HIV-1 coreceptor tropism specifically for subtype A sequences.
• To address the limitations of current models in accurately predicting tropism for non-B HIV-1 subtypes.

Main Methods:

• Development of a novel model named SCOTCH (Stacked Classifier for Coreceptor Tropism prediction in HIV).
• SCOTCH utilizes a stacking approach combining multiple classifier ensembles.
• Model performance was evaluated against existing tropism prediction tools.

Main Results:

• SCOTCH demonstrates significantly improved prediction performance for subtype A HIV-1 sequences compared to existing models.
• The model shows superior accuracy, particularly at low false positive rates (0.05-0.2), aligning with clinical guidelines.
• SCOTCH achieved better partial area under the curves and diagnostic odds ratios.

Conclusions:

• SCOTCH provides a reliable tool for predicting coreceptor tropism in subtype A HIV-1.
• This improved prediction accuracy can enhance the selection of appropriate antiretroviral therapy regimens.
• The SCOTCH model is accessible for research and clinical application.