PERGA

PERGA assembles high-throughput short sequencing reads into contigs and scaffolds for de novo genome assembly by leveraging paired-end information to resolve repeats and branching ambiguities.

Key Features:

• Greedy-like prediction strategy: Assembles sequence reads into contigs and scaffolds using a greedy-like prediction that leverages paired-end reads and varying overlap sizes (Omax to Omin) to resolve gaps and branches.
• Machine learning-based decision model: Uses a machine learning decision model that exploits branch features to choose the correct extension path, reported to achieve 99.7% correctness.
• Handling of tandem repeats: Detects different copies of tandem repeats to enable more accurate extension through adjacent repetitive sequences that commonly fragment assemblies.
• Look-ahead extension approach: When immediate extension is uncertain, attempts all feasible extensions and subsequently identifies the correct path to improve assembly continuity and accuracy.

Scientific Applications:

• De novo assembly of diverse genomes: Applied to real Illumina datasets and simulated data across genome sizes from small bacterial genomes to large human chromosomes to construct longer and more accurate contigs and scaffolds.

Methodology:

Computational methods include a greedy-like prediction strategy using overlaps (Omax to Omin) and paired-end reads; a machine learning decision model using branch features (99.7% correctness reported); detection of distinct tandem repeat copies; and a look-ahead procedure that attempts all feasible extensions and selects the correct path, producing contigs and scaffolds.