EMIRGE

EMIRGE reconstructs near–full-length small subunit (SSU) ribosomal RNA genes, notably 16S rRNA, from short-read sequencing data and estimates relative abundances of the reconstructed sequences for microbial community profiling.

Key Features:

• Full-length SSU (16S rRNA) reconstruction: Reconstructs near–full-length ribosomal gene sequences from short-read data.
• Abundance estimation: Estimates sequence abundances for reconstructed SSU sequences within samples.
• Expectation-Maximization iterative refinement: Adapts the Expectation-Maximization (EM) algorithm to iteratively refine abundance estimates and read-origin probabilities.
• Probabilistic read assignment: Calculates the probability that each sequencing read originated from specific reference sequences.
• Consensus sequence correction: Uses read-origin probabilities to correct consensus sequences for each reference SSU sequence each iteration.
• Read mapping with Bowtie: Maps reads to reference sequences using Bowtie alignments, with design support for alternative mappers.
• Convergence behavior: Repeats mapping and estimation until convergence, typically after 40–80 iterations.
• Sensitivity and taxonomic resolution: Detects low-abundance organisms (as low as ~0.01% of a community) and resolves taxa including members of uncultivated phyla.

Scientific Applications:

• Microbial community profiling: Reconstructs and quantifies SSU rRNA sequences to profile community composition from short-read datasets.
• Diversity and rare taxa detection: Identifies rare and low-abundance organisms and captures diversity across cultivated and uncultivated phyla.
• Environmental perturbation studies: Resolves shifts in community composition under perturbations, including transitions such as acetate amendment-driven iron reduction to sulfate reduction.
• Comparative community analysis: Differentiates community compositions across environmental states and large datasets with reproducible results across technical replicates.
• Population dynamics: Detects taxa-level responses to selective pressures and temporal changes within microbial communities.

Methodology:

EMIRGE takes short-read sequencing data, maps reads to reference sequences using Bowtie alignments, and applies an adapted Expectation-Maximization algorithm that estimates expected abundances of SSU sequences and probabilities that each read originated from particular references; it uses those probabilities to correct consensus sequences and repeats mapping and estimation until convergence, typically after 40–80 iterations.