Simka

Simka computes k-mer–based ecological distances between metagenomic datasets to enable de novo comparative metagenomics without relying on taxonomic or functional assignation.

Key Features:

• K-mer spectrum representation: Represents each metagenomic dataset as a k-mer spectrum, replacing species counts with k-mer counts.
• Ecological distance computation: Computes a variety of classical ecological distances, including both qualitative and quantitative measures, between k-mer spectra.
• Parallel k-mer counting: Implements a parallel k-mer counting strategy that processes multiple datasets simultaneously to improve computational efficiency.
• Scalability: Scales to large-scale metagenomic projects, demonstrated on hundreds of samples and tens of billions of reads (e.g., 690 samples and >32 billion reads processed in a few hours).
• Reference-free comparison: Compares entire sequence sets without taxonomic or functional assignation, enabling analysis of sequences not associated with known organisms.
• Concordance with alignment and profiling: K-mer level analyses show high correlation with de novo comparison techniques based on exhaustive all-versus-all sequence alignment and with taxonomic profiling.
• Validation on public datasets: Validated on Human Microbiome Project datasets and shown to capture the underlying biological structure of microbial communities.

Scientific Applications:

• Comparative metagenomics: Compare microbial community composition across diverse environmental or host-associated conditions using de novo, reference-free metrics.
• Biodiversity assessment: Extract biodiversity knowledge from complex environmental samples without relying on reference databases.
• Large-scale community analysis: Perform large-scale comparative analyses across hundreds of metagenomic samples for population-level or cohort studies.
• Method benchmarking: Benchmark and validate de novo comparison approaches against exhaustive alignment-based and taxonomic profiling methods.

Methodology:

Each dataset is represented as a k-mer spectrum; k-mers are counted using a parallel counting strategy across datasets; classical ecological distances (qualitative and quantitative) are computed between k-mer spectra, with k-mer counts replacing species counts.