AntiHIV-Pred

AntiHIV-Pred predicts anti-HIV activity and related biological activities of small molecules using QSAR and PASS-based in silico models to support discovery and prioritization of anti-HIV therapeutics.

Key Features:

• ChEMBL dataset: Training data comprise over 50,000 experimental records of anti-retroviral agents extracted from the ChEMBL database.
• Model construction (GUSAR): Approximately 7,000 molecules known to inhibit five key HIV-1 proteins were used to build regression and classification QSAR models with GUSAR software.
• QSAR performance: QSAR models achieved average R² = 0.95 and Q² = 0.72 during validation.
• PASS activity prediction: The PASS program predicts 81 different biological activities related to HIV-associated comorbidities with a mean accuracy of 92%.

Scientific Applications:

• Drug discovery prioritization: Prioritizes candidate small molecules for experimental validation in anti-HIV therapeutic development.
• Identification of candidates for HIV and comorbidities: Identifies promising compounds for HIV treatment and management of HIV-associated comorbid conditions based on predicted activities.
• Reduction of experimental burden: Reduces reliance on costly and time-intensive laboratory experiments by providing in silico activity estimates.

Methodology:

Experimental records were extracted from ChEMBL; regression and classification QSAR models were constructed with GUSAR using ~7,000 molecules inhibiting five key HIV-1 proteins and validated (average R² = 0.95, Q² = 0.72); biological activity spectra were predicted with the PASS program for 81 activities (mean accuracy 92%).