ConnectedReads

ConnectedReads performs machine-learning-optimized whole-read assembly of short-read next-generation sequencing (NGS) data to generate assembled contigs that enhance structural-variant (SV) discovery in regions of high divergence and N-gap regions.

Key Features:

• Whole-Read Assembly Workflow: Implements a whole-read assembly workflow using unsupervised graph mining algorithms executed on the Apache Spark large-scale data processing platform.
• Utilization of Short-Read Data: Fully exploits information from short-read NGS data to produce assembled contigs that represent more complete genomic sequences than individual reads.
• Enhanced Structural Variant Discovery: Improves resolution of SV discovery, especially in regions with high diversity relative to reference genomes and in N-gap regions.
• Population-Scale Analysis: Enables cost-effective, large-scale investigation of genetic variation across populations through scalable assembly-based analysis.

Scientific Applications:

• Clinical genomics: Detection and characterization of complex genetic variants to inform studies of genetic diseases.
• Population genetics: Large-scale analysis of genetic diversity and variant frequency across cohorts.
• Evolutionary studies: Characterization of sequence divergence and structural variation relevant to comparative and evolutionary analyses.
• Reference-aware variant discovery: Improved variant detection in contexts with incomplete or divergent reference genomes and N-gap regions.

Methodology:

ConnectedReads integrates machine-learning optimization with unsupervised graph mining algorithms to perform whole-read assembly on short-read NGS data within the Apache Spark framework to generate assembled contigs and improve structural-variant discovery.