Mapsembler, targeted and micro assembly of large NGS datasets on a desktop computer

Area of Science:

• Genomics
• Bioinformatics
• Computational Biology

Background:

• Next-generation sequencing generates massive datasets of short DNA fragments.
• De novo assemblers, while powerful, demand significant computational resources (memory and time).
• Targeted analysis of specific genomic regions offers an alternative to complete genome assembly for answering biological questions.

Purpose of the Study:

• To develop a computational method for analyzing large-scale sequencing data efficiently.
• To enable targeted de novo discovery around specific genomic regions of interest.
• To reduce the computational burden associated with whole genome or transcriptome assembly.

Main Methods:

• Development of Mapsembler, an iterative micro and targeted assembler.
• Implementation of algorithms for retrieving approximate sequence occurrences from reads.
• Construction of an extension graph to represent sequence context and structure.

Main Results:

• Mapsembler processes large sequencing datasets on commodity hardware.
• The software can identify and assemble regions of interest as sequences or graphs.
• Successfully retrieved known human breast cancer candidate fusion genes and identified novel ones.

Conclusions:

• Mapsembler is the first software for de novo discovery of repeats, SNPs, exon skipping, and gene fusions directly from raw sequencing reads.
• The localized indexing approach results in a negligible memory footprint.
• Mapsembler is freely available under the CeCILL license.