GPTomics/bio-microbiome-diversity-analysis

Version Compatibility

Reference examples tested with: R stats (base), ggplot2 3.5+, phyloseq 1.46+, scanpy 1.10+, vegan 2.6+

Before using code patterns, verify installed versions match. If versions differ:

• R: packageVersion('<pkg>') then ?function_name to verify parameters

If code throws ImportError, AttributeError, or TypeError, introspect the installed package and adapt the example to match the actual API rather than retrying.

Diversity Analysis

"Compare microbial diversity across my samples" -> Calculate alpha diversity (within-sample richness/evenness) and beta diversity (between-sample dissimilarity) to test for community composition differences across groups.

• R: phyloseq::estimate_richness() for alpha, phyloseq::ordinate() for beta
• R: vegan::adonis2() for PERMANOVA testing

Create phyloseq Object

library(phyloseq)
library(vegan)
library(ggplot2)

seqtab <- readRDS('seqtab_nochim.rds')
taxa <- readRDS('taxa.rds')
metadata <- read.csv('sample_metadata.csv', row.names = 1)

ps <- phyloseq(otu_table(seqtab, taxa_are_rows = FALSE),
               tax_table(taxa),
               sample_data(metadata))
taxa_names(ps) <- paste0('ASV', seq(ntaxa(ps)))

Alpha Diversity

# Calculate multiple metrics
alpha_div <- estimate_richness(ps, measures = c('Observed', 'Chao1', 'Shannon', 'Simpson'))
alpha_div$SampleID <- rownames(alpha_div)
alpha_div <- merge(alpha_div, sample_data(ps), by = 'row.names')

# Statistical test
kruskal.test(Shannon ~ Group, data = alpha_div)

# Pairwise comparisons
pairwise.wilcox.test(alpha_div$Shannon, alpha_div$Group, p.adjust.method = 'BH')

Alpha Diversity Plots

plot_richness(ps, x = 'Group', measures = c('Observed', 'Shannon')) +
    geom_boxplot() +
    theme_minimal()

# Custom plot
ggplot(alpha_div, aes(x = Group, y = Shannon, fill = Group)) +
    geom_boxplot() +
    geom_jitter(width = 0.2, alpha = 0.5) +
    theme_minimal() +
    labs(y = 'Shannon Diversity Index')

Faith's Phylogenetic Diversity

Goal: Calculate phylogenetic alpha diversity (Faith's PD) from ASV data by building a de novo phylogeny and summing branch lengths.

Approach: Align ASV sequences with DECIPHER, construct a neighbor-joining tree with phangorn, root at midpoint, and compute PD using picante.

library(picante)

# Requires phylogenetic tree in phyloseq object
# Build tree from ASV sequences
library(DECIPHER)
library(phangorn)

seqs <- refseq(ps)
alignment <- AlignSeqs(seqs, anchor = NA)
phang_align <- phyDat(as(alignment, 'matrix'), type = 'DNA')
dm <- dist.ml(phang_align)
tree <- NJ(dm)
tree <- midpoint(tree)
phy_tree(ps) <- tree

# Calculate Faith's PD
otu_mat <- as.matrix(t(otu_table(ps)))
faith_pd <- pd(otu_mat, phy_tree(ps), include.root = TRUE)
alpha_div$PD <- faith_pd$PD

Rarefaction Curves

# Check if sequencing depth is adequate
rarecurve_data <- vegan::rarecurve(t(otu_table(ps)), step = 100, sample = min(sample_sums(ps)))

# ggplot version with ggrare (install from GitHub)
# devtools::install_github('gauravsk/ranacapa')
library(ranacapa)
p_rare <- ggrare(ps, step = 100, color = 'Group', se = FALSE)
p_rare + theme_minimal() + labs(title = 'Rarefaction Curves')

Rarefaction

# Check sequencing depth
sample_sums(ps)

# Rarefy to minimum depth
ps_rarefied <- rarefy_even_depth(ps, sample.size = min(sample_sums(ps)),
                                  rngseed = 42, replace = FALSE)

Beta Diversity

# Calculate distance matrices
bray <- phyloseq::distance(ps, method = 'bray')       # Bray-Curtis
jaccard <- phyloseq::distance(ps, method = 'jaccard') # Jaccard
unifrac <- UniFrac(ps, weighted = TRUE)               # Weighted UniFrac (requires tree)

# Ordination
ord_bray <- ordinate(ps, method = 'PCoA', distance = bray)

# Plot
plot_ordination(ps, ord_bray, color = 'Group') +
    stat_ellipse(level = 0.95) +
    theme_minimal()

PERMANOVA

# Test for group differences
metadata <- data.frame(sample_data(ps))
permanova_result <- adonis2(bray ~ Group, data = metadata, permutations = 999)
permanova_result

# With covariates
adonis2(bray ~ Group + Age + Sex, data = metadata, permutations = 999)

Beta Dispersion

# Test homogeneity of dispersions (assumption of PERMANOVA)
beta_disp <- betadisper(bray, metadata$Group)
permutest(beta_disp)
plot(beta_disp)

NMDS Ordination

ord_nmds <- ordinate(ps, method = 'NMDS', distance = bray)

# Check stress
ord_nmds$stress  # Should be < 0.2

plot_ordination(ps, ord_nmds, color = 'Group') +
    theme_minimal()

Distance Metrics Comparison

| Metric | Type | Considers Abundance | Phylogeny | |--------|------|---------------------|-----------| | Bray-Curtis | Quantitative | Yes | No | | Jaccard | Binary | No | No | | UniFrac (unweighted) | Binary | No | Yes | | UniFrac (weighted) | Quantitative | Yes | Yes |

Related Skills

• amplicon-processing - Generate ASV table
• differential-abundance - Identify taxa driving differences
• data-visualization/ggplot2-fundamentals - Custom diversity plots
• ecological-genomics/biodiversity-metrics - Hill number coverage-based rarefaction for ecological data
• ecological-genomics/community-ecology - Constrained ordination and indicator species