Swarm

Swarm clusters amplicon sequences into operational taxonomic units (OTUs) using an iterative local threshold approach to provide fine-scale, input-order-independent OTU delineation for amplicon-based studies.

Key Features:

• Iterative local clustering: Clusters nearly identical amplicons using an iterative local threshold rather than a global similarity threshold.
• Abundance- and structure-aware refinement: Leverages internal cluster structure and abundance information to refine OTUs.
• Input-order independence: Produces OTUs that are independent of input order and reduces reliance on arbitrary global clustering parameters.
• d = 1 algorithm (Swarm v2): Implements a novel algorithm for d = 1 that enables linear scaling of computation time with increasing data volumes.
• Fastidious option (Swarm v2): Minimizes under-grouping by integrating low-abundance OTUs such as singletons and doubletons into larger clusters.
• Integrated clustering and breaking: Directly integrates clustering and breaking phases for consolidated processing.
• Dereplication (d = 0): Supports dereplication of sequencing reads with d = 0.
• FASTA output: Writes OTU representatives in FASTA format.
• OTU network visualization: Plots individual OTUs as two-dimensional networks.

Scientific Applications:

• Amplicon-based studies: Generation of operational taxonomic units (OTUs) from amplicon sequencing data for biodiversity and community analyses.
• Fine-scale molecular OTU delineation: Delineation of high-resolution molecular OTUs without arbitrary global thresholds.
• Large dataset processing: Scalable clustering applicable to large sequencing datasets due to the d = 1 linear-scaling algorithm.
• OTU structure visualization and dereplication: Visualization of OTU internal networks and dereplication of reads for downstream analyses.

Methodology:

Uses an iterative local threshold clustering approach that leverages internal cluster structure and abundance information, implements a linear-scaling algorithm for d = 1, offers a "fastidious" option to merge low-abundance OTUs, integrates clustering and breaking phases, supports dereplication at d = 0, outputs OTU representatives in FASTA, and plots OTUs as two-dimensional networks.