MetaPhlAn

MetaPhlAn relies on a curated catalog of universal single-copy marker genes that are uniquely specific to individual microbial clades. Unlike k-mer approaches, this strategy avoids ambiguous read assignments by targeting genomic regions that are both ubiquitous within a clade and absent from others. The database is built from sequenced isolate genomes, with markers selected through a rigorous pipeline that identifies genes present in ≥90% of target clade genomes and absent in outgroup genomes. This gene-centric rather than whole-genome approach dramatically reduces computational overhead while maintaining high taxonomic resolution.

MetaPhlAn achieves strain-level and species-level resolution by mapping shotgun metagenomic reads directly to its marker gene database using Bowtie2. The algorithm normalizes mapped read counts by marker gene length and single-copy status to compute relative abundance for each detected clade. This mapping-first approach enables detection of organisms present at abundances as low as 0.01%, making it suitable for low-biomass samples. The output is a community composition table directly comparable across samples without the need for assembly or binning.

The companion tool StrainPhlAn extends MetaPhlAn's capabilities to sub-species strain tracking. It reconstructs consensus sequences for dominant species by extracting and concatenating species-specific marker genes, then performs multiple sequence alignment across samples. This enables phylogenetic placement of strains, tracking transmission events, and distinguishing closely related variants within a species. StrainPhlAn is widely used in outbreak investigations and longitudinal microbiome studies to monitor strain persistence and replacement dynamics.

MetaPhlAn bypasses the computationally expensive steps of metagenomic assembly and binning by operating directly on raw sequencing reads. The marker gene database is compact, and read mapping with Bowtie2 is highly optimized. This design enables profiling of hundreds of samples in hours on standard hardware, compared to assembly-based workflows that require days and substantial memory. The trade-off is that MetaPhlAn does not recover novel genomes or functional gene content, focusing exclusively on taxonomic composition.

MetaPhlAn integrates with HUMAnN (HMP Unified Metabolic Analysis Network) to provide functional profiling. While MetaPhlAn identifies which organisms are present, HUMAnN maps reads to pangenome and pathway databases to quantify gene families and metabolic pathways. The integration uses MetaPhlAn's abundance estimates to perform species-level functional contribution analysis, attributing specific metabolic functions to their source organisms. This layered approach answers both 'who is there?' and 'what are they doing?' in a single workflow.

MetaPhlAn maintains strict database versioning to ensure computational reproducibility across studies. Each release (e.g., MetaPhlAn 4.0) includes an expanded set of marker genes derived from an updated collection of reference genomes. The marker selection algorithm is deterministic and documented, allowing researchers to reproduce exact abundance profiles by specifying the database version. This versioning is critical for longitudinal studies and meta-analyses where consistent taxonomic definitions must be maintained across time points and cohorts.

The foundational profiling strategy MetaPhlAn employs. Instead of whole-genome alignment, it identifies and quantifies a predefined set of clade-specific, single-copy, universal marker genes. This dramatically reduces computational overhead and avoids biases from horizontal gene transfer or variable genome sizes. The approach assumes that the count of these markers is a direct proxy for organismal abundance, enabling rapid, accurate species-level relative abundance estimation.

The broader computational goal MetaPhlAn serves: characterizing the microbial community structure of a metagenomic sample. Profiling answers 'who is there and in what proportion?' Key distinctions include:

• Marker-gene methods (MetaPhlAn): Fast, database-dependent, focus on known clades.
• k-mer methods (Kraken2): Exact-match queries against entire genomes, memory-intensive.
• Binning methods: Assemble genomes first, then classify, enabling discovery of novel organisms.

MetaPhlAn's ability to distinguish genetic variants below the species rank using strain-specific marker genes. This is critical for tracking pathogen outbreaks, understanding functional differences within a species (e.g., pathogenic E. coli O157:H7 vs. commensal K-12), and analyzing subspecies population dynamics. MetaPhlAn 4 achieves this by incorporating a pan-genome database, capturing the core and accessory gene diversity needed for precise strain identification.

The untargeted sequencing approach that generates the input data for MetaPhlAn. Unlike 16S rRNA amplicon sequencing, which targets a single gene, shotgun metagenomics fragments and sequences all genomic DNA in a sample. This provides the comprehensive genomic coverage necessary for MetaPhlAn to query its database of universal single-copy markers, enabling functional profiling and strain-level resolution that amplicon methods cannot achieve.

A conservative classification strategy used by tools like Kraken2, contrasting with MetaPhlAn's direct marker-to-clade mapping. When a read matches multiple reference genomes, the LCA algorithm assigns it to the deepest taxonomic node common to all matches. This minimizes false-positive species calls but often leaves reads classified at higher ranks (e.g., genus). MetaPhlAn's clade-specific markers avoid this ambiguity by design, providing definitive species-level assignments.

The downstream analysis often paired with MetaPhlAn's taxonomic output. While MetaPhlAn identifies who is present, tools like HUMAnN 3 use its abundance profiles to stratify metabolic pathway contributions by organism. This layered approach answers 'what are they doing?' by mapping sequencing reads to gene families (e.g., UniRef90) and pathways (e.g., MetaCyc), revealing the community's collective metabolic potential.