Minerva

Minerva deconvolves Linked-Read sequencing data tagged with 3' barcodes (UIDs) into clusters that represent individual long DNA fragments to improve metagenomic analysis.

Key Features:

• Alignment- and Reference-Free: Performs deconvolution without alignment to reference genomes, enabling analysis when references are incomplete or absent.
• Graph-Based Algorithm: Uses a graph-based algorithm that leverages connectivity patterns among barcoded reads to cluster reads sharing a single 3' barcode into fragment-specific groups.
• Linked-Read Support: Tailored for Linked-Read technologies such as the 10x Chromium system that use microfluidic barcoding of short reads sequenced on short-read platforms (e.g., Illumina).
• Robustness to Sparse Coverage and Barcode Ambiguity: Resolves cases where a single barcode corresponds to multiple fragments and operates under sparse coverage and lack of read order information.
• Enhances Downstream Analyses: Produces deconvolved clusters that improve the specificity of downstream taxonomic assignment and k-mer-based clustering approaches.
• Leverages Linked-Read Advantages: Exploits the cost-effective, lower-input, and Illumina-comparable error profile of Linked-Read data relative to long-read technologies.

Scientific Applications:

• Improved Taxonomic Assignment: Increases specificity of taxonomic assignments in complex microbial communities by clustering reads into their originating long DNA fragments.
• Enhanced k-mer-Based Clustering and Assembly Support: Improves performance of k-mer-based clustering methods used for genome assembly and assessing genetic diversity in metagenomic samples.

Methodology:

Minerva applies an alignment- and reference-free, graph-based algorithm that leverages connectivity among barcoded reads to deconvolve reads sharing a single 3' barcode into clusters corresponding to distinct long DNA fragments, handling sparse coverage and barcode ambiguity in Linked-Read data (e.g., 10x Chromium).