PredictHaplo

PredictHaplo reconstructs viral haplotypes from next-generation sequencing (NGS) and short deep sequencing reads to identify HIV variants, including potential drug-resistant mutants.

Key Features:

• Haplotype Reconstruction: Reconstructs haplotypes from short deep sequencing reads and addresses the absence of direct pairwise similarity between non-overlapping sequences.
• Statistical Framework: Frames haplotype inference as a nonstandard clustering problem and employs a Dirichlet Process Mixture Model (DPMM).
• Sequential Updating: Sequentially updates prior information through successive local analyses to refine haplotype estimates.
• Validation: Performance tested and validated on both simulated and real sequencing datasets.

Scientific Applications:

• HIV intra-host diversity analysis: Identifies diverse HIV haplotypes within a patient from NGS data to characterize intra-host viral diversity.
• Drug resistance detection: Detects potentially drug-resistant HIV mutants to inform antiretroviral treatment considerations.

Methodology:

Uses a Dirichlet Process Mixture Model (DPMM) to perform nonstandard clustering on short deep sequencing reads, sequentially updates priors via successive local analyses, and reconstructs haplotypes despite non-overlapping reads; validated on simulated and real sequencing datasets.