Metagenomics: Microbiome Ecology, Shotgun Taxonomic Profiling & Uncultivated Genomes
Metagenomics decodes the taxonomic diversity, functional resistome, and metabolic potential of complex uncultured microbial communities inhabiting the human body, soils, and oceans. By bypassing laboratory culture bottlenecks, high-throughput metagenomics reconstructs complete Metagenome-Assembled Genomes (MAGs) and reveals host-microbe metabolic crosstalk.
🦠 Module 1: Biological Foundations & The Uncultured Majority
Microbial EcologyLess than 1% of environmental and host-associated microorganisms can be cultured in pure laboratory media—a dilemma famously known as the Great Plate Count Anomaly. Metagenomics circumvents cultivation by extracting high-molecular-weight environmental DNA directly from biological specimens (stool, soil, ocean water, saliva) and interrogating the pooled genomes of all co-existing microbes simultaneously.
The 16S rRNA Molecular Clock
The prokaryotic 16S ribosomal RNA gene (~1,540 bp) is universal across Bacteria and Archaea. It contains 9 highly conserved primer-binding regions (C1–C9) flanking 9 hypervariable loops (V1–V9). Primers targeting V3–V4 or V4 enable amplicon-based phylogenetic classification.
Alpha & Beta Ecological Diversity
Community structure is evaluated across two axes:
• Alpha Diversity: Within-sample richness and evenness (Shannon Index, Simpson, Chao1).
• Beta Diversity: Between-sample compositional dissimilarity (Bray-Curtis, Weighted/Unweighted UniFrac) visualized via PCoA ordination.
The Functional Resistome & BGCs
Beyond taxonomy, metagenomic assemblies uncover antimicrobial resistance genes (ARGs) conferring multidrug resistance (beta-lactamases, efflux pumps) and Biosynthetic Gene Clusters (BGCs) producing novel secondary metabolite antibiotics.
🔬 Module 2: Marker-Gene Amplicon vs Shotgun Metagenomics
Methodological Comparison| Paradigm | Target DNA & Strategy | Taxonomic Resolution | Functional Profiling | Sequencing Depth Required |
|---|---|---|---|---|
| 16S / 18S / ITS Amplicon | PCR amplification of targeted hypervariable locus using universal degenerate primers. | Genus level (rarely species-level due to 16S conservation). | Indirect prediction only (PICRUSt2). | Low (50,000 – 100,000 reads per sample). |
| Whole Metagenome Shotgun (WMS) | Untargeted physical shearing of total environmental DNA followed by high-depth shotgun sequencing. | Species and Strain level. | Direct gene and pathway quantification (HUMAnN3). | High (10 – 30 million paired-end reads per sample). |
| Long-Read Metagenomics (HiFi / ONT) | Direct sequencing of 10–50 kb native fragments to bridge structural operons and circular plasmids. | Strain & closed circular chromosomes. | Complete operons & mobile genetic elements. | Moderate (1 – 3 million long reads per sample). |
🖼️ Module 3: Visual Flowchart & MAG Binning Architecture
Scientific Diagram💻 Module 4: Computational Pipelines & Production Code
DADA2 & Shotgun MAGs# Production R Script: DADA2 ASV Denoising, Chimera Purging & SILVA Taxonomy
library(dada2)
path <- "./raw_16s_fastq"
fnFs <- sort(list.files(path, pattern="_R1_001.fastq.gz", full.names = TRUE))
fnRs <- sort(list.files(path, pattern="_R2_001.fastq.gz", full.names = TRUE))
# 1. Filter and Trim reads (truncate at Phred Q < 2 drops)
filtFs <- file.path(path, "filtered", basename(fnFs))
filtRs <- file.path(path, "filtered", basename(fnRs))
out <- filterAndTrim(fnFs, filtFs, fnRs, filtRs, truncLen=c(240,160), maxN=0, maxEE=c(2,2), truncQ=2, rm.phix=TRUE, multithread=16)
# 2. Learn Parametric Error Rates from biological reads
errF <- learnErrors(filtFs, multithread=16)
errR <- learnErrors(filtRs, multithread=16)
# 3. Sample Inference & Dereplication
dadaFs <- dada(filtFs, err=errF, multithread=16)
dadaRs <- dada(filtRs, err=errR, multithread=16)
# 4. Merge Paired Reads into Contigs (requiring >= 12 bp overlap)
mergers <- mergePairs(dadaFs, filtFs, dadaRs, filtRs, minOverlap=12)
# 5. Construct Sequence Table & Remove PCR Chimeras
seqtab <- makeSequenceTable(mergers)
seqtab.nochim <- removeBimeraDenovo(seqtab, method="consensus", multithread=16)
# 6. Assign Taxonomy using naive Bayesian classifier against SILVA 138.2
taxa <- assignTaxonomy(seqtab.nochim, "silva_nr99_v138.2_train_set.fa.gz", multithread=16)
saveRDS(seqtab.nochim, "dada2_asv_matrix.rds")
saveRDS(taxa, "dada2_taxonomy.rds")#!/usr/bin/env bash
set -euo pipefail
# 1. Purge Host Human DNA with Bowtie2
bowtie2 -p 16 -x /genomes/GRCh38/bowtie2_index/GRCh38 \
-1 sample_R1.fastq.gz -2 sample_R2.fastq.gz --un-conc-gz sample_microbiome_R%.fq.gz > /dev/null
# 2. Ultra-Fast Exact k-mer Taxonomic Classification with Kraken2 & Bracken
kraken2 --db /databases/kraken2_standard/ --threads 16 \
--paired sample_microbiome_R1.fq.gz sample_microbiome_R2.fq.gz \
--report sample_kraken2_report.txt --output sample_kraken2_output.txt
bracken -d /databases/kraken2_standard/ -i sample_kraken2_report.txt -o sample_bracken_species.txt -r 150 -l S
# 3. De Novo Metagenome Assembly with MEGAHIT
megahit -1 sample_microbiome_R1.fq.gz -2 sample_microbiome_R2.fq.gz \
-o megahit_assembly/ --min-contig-len 1500 -t 16
# 4. Map Reads to Contigs for Differential Coverage Estimation
bwa-mem2 index megahit_assembly/final.contigs.fa
bwa-mem2 mem -t 16 megahit_assembly/final.contigs.fa sample_microbiome_R1.fq.gz sample_microbiome_R2.fq.gz | \
samtools sort -@ 8 -o contig_coverage.bam -
samtools index contig_coverage.bam
jgi_summarize_bam_contig_depths --outputDepth depth.txt contig_coverage.bam
# 5. Metagenome Binning with MetaBAT2 & Quality Verification with CheckM2
metabat2 -i megahit_assembly/final.contigs.fa -a depth.txt -o ./metabat_bins/bin -m 1500 -t 16
checkm2 predict --threads 16 --input ./metabat_bins/ -x fa --output-directory ./checkm2_qc/🏥 Module 5: Real-World Clinical Case Study
Gut Microbiome DysbiosisCase: Metagenomic Deconstruction of Severe Dysbiosis in Active Crohn's Disease
Clinical Problem: Inflammatory Bowel Disease (IBD) patients experience recurrent mucosal inflammation. Longitudinal deep shotgun metagenomics (30 million reads/sample) was conducted across a cohort of 60 Crohn’s disease patients versus 40 age-matched healthy household controls.
Expanded Pathobionts: Adherent-Invasive Escherichia coli (AIEC, log2FC = +5.34, padj = 1.4e-16)
Functional Pathway Depletion: Butyrate Fermentation via Butyryl-CoA:Acetate CoA-Transferase (MetaCyc CENTFERM-PWY, p < 10^-12)
Resistome Findings: Marked increase in beta-lactamase (blaTEM-1) and aminoglycoside resistance cassettes.
Translational Resolution: Loss of butyrate-producing anti-inflammatory taxa was linked to intestinal barrier breakdown. Stratification via MAG functional gene profiles predicted which patients would fail anti-TNF biologic therapy prior to drug administration.
📚 Complete 13-Chapter Metagenomics Curriculum
1: The Microbiome Landscape: Amplicon vs Shotgun Metagenomics →
16S/18S/ITS primer choice, mock communities, and contamination controls.
2: Amplicon Read Processing & DADA2 Denoising →
ASV inference, sample dereplication, and chimera filtering.
3: Taxonomic Assignment & QIIME 2 Workflows →
Naive Bayes classifiers, SILVA, Greengenes2, and phylogenetic trees.
4: Alpha & Beta Diversity Statistics in R →
Shannon, Chao1, Bray-Curtis, UniFrac, and PERMANOVA (adonis2).
5: Differential Abundance Testing (ANCOM-BC2 & MaAsLin2) →
Handling microbiome compositionality, zero-inflation, and false discovery.
6: Shotgun Metagenomics QC & Host Decontamination →
Purging human host reads using kneaddata, Bowtie2, and BWA-MEM2.
7: Ultra-Fast Taxonomic Profiling with Kraken2 & Bracken →
Exact k-mer mapping and Bayesian abundance re-estimation.
8: Clade-Specific Marker Gene Profiling (MetaPhlAn 4) →
High-precision species/strain classification with SGB catalog.
9: Metabolic Pathway Reconstruction (HUMAnN3) →
UniRef90 gene family abundance and MetaCyc pathway coverage.
10: De Novo Metagenome Assembly (metaSPAdes & MEGAHIT) →
De Bruijn graph assembly across uneven sequencing depth.
11: Metagenome-Assembled Genomes (MAGs): Binning with MetaBAT2 →
Multi-sample co-abundance and DAS Tool ensemble refinement.
12: MAG Quality Assessment (CheckM2) & GTDB-Tk →
Machine learning completeness estimation and phylogenetic trees.
13: Antimicrobial Resistance (AMR) & Viral Metagenomics →
CARD RGI, AMRFinderPlus, and viral contig identification (VirSorter2).
📦 Module 6: Essential Software Directory & Official Links
Verified Tool StackDADA2
Amplicon ASVsFast, accurate sample inference from amplicon data with exact sequence variant resolution.
DADA2 Official Documentation →Kraken 2 & Bracken
Taxonomic ClassifierUltra-fast k-mer taxonomic classification with Bayesian abundance re-estimation at the species level.
JHU Kraken 2 Portal →MetaBAT 2
MAG BinningMetagenome Binning based on tetranucleotide frequencies and multi-sample differential coverage.
Berkeley Lab MetaBAT →CheckM2
Quality BenchmarkingMachine learning-based prediction of completeness and contamination for uncultivated bacterial and archaeal genomes.
CheckM2 GitHub →HUMAnN 3
Functional ProfilingHuttenhower Lab tool for accurately profiling the presence and abundance of microbial pathways and functional gene families.
Harvard HUMAnN 3 →CARD Resistance Gene Identifier
Resistome • AMRThe Comprehensive Antibiotic Resistance Database software for predicting resistomes from genome and metagenome contigs.
McMaster CARD RGI →⚠️ Module 7: Bioinformatic Pitfalls & Quality Control Standards
Best Practices1. Low-Biomass "Kitome" Contamination
In low-biomass specimens (e.g. lung, blood, placenta), trace microbial DNA contaminating commercial DNA extraction spin columns (e.g. *Ralstonia*, *Pseudomonas*) can dominate sequencing reads. Always sequence negative blank extraction controls.
2. Chimeric MAG Assembly
Closely related strains sharing high sequence homology frequently cross-assemble into single chimeric contigs. Always verify MAG purity with single-copy marker genes in CheckM2 (<5% contamination required).
3. Compositionality & False Correlation
Sequencing generates relative abundance data constrained to a constant sum ($100\%$). Expansion of one taxon artificially suppresses the relative fraction of all other taxa. Never use standard Pearson correlation; use log-ratio models (ANCOM-BC2, MaAsLin2).
📝 Module 8: Self-Assessment Exercises & Practice Datasets
Hands-on MasteryQ1: Why have Amplicon Sequence Variants (ASVs) largely replaced 97% Operational Taxonomic Units (OTUs)?
Answer: 97% OTUs cluster sequences by an arbitrary sequence identity threshold, often collapsing distinct biological bacterial strains into single bins while obscuring true biological variation. In contrast, DADA2 models the experimental error rate of Illumina sequencers to distinguish true biological variants from sequencing errors down to single-nucleotide differences. Furthermore, ASVs are exact sequences that can be directly compared across independent studies and datasets without re-clustering.
Q2: What is the biological and mathematical principle behind differential coverage binning in MetaBAT2?
Answer: All contigs that originate from the same bacterial chromosome must share the exact same physical abundance within a given sample. Across multiple independent biological samples or timepoints, contigs belonging to the same organism will covary in abundance together. MetaBAT2 pairs this differential coverage covariance with tetranucleotide frequency (TNF) genomic signatures to accurately group contigs into discrete organismal genomes (MAGs).
Q3: Why must human host reads be purged before executing de novo metagenome assembly?
Answer: In clinical specimens (e.g. mucosal swabs, saliva, biopsy tissue), host human DNA often accounts for 90–99% of raw sequencing reads. Attempting to assemble raw files wastes massive computational memory on the 3.05 Gb human genome, creates spurious human-microbe chimeric contigs, and violates patient privacy standards (HIPAA/GDPR) by retaining identifiable human germline sequence data.