4. Community structure (PCoA)

Analyse community composition using Bray–Curtis distances and visualise patterns with a Principal Coordinates Analysis (PCoA).

Normalise counts to relative abundances

First, convert ASV counts to relative abundances per sample:

norm_counts <- t(t(merged_df$counts) / colSums(merged_df$counts, na.rm = TRUE))

Calculate Bray–Curtis distances

Compute pairwise Bray–Curtis distances (beta-diversity):

bray_dist <- as.matrix(vegdist(t(norm_counts), method = "bray"))

Run PCoA

Perform Principal Coordinates Analysis on the distance matrix:

pcoa <- pcoa(bray_dist, correction = "cailliez")

Plot PCoA results

We plot samples in PCoA space using the first two axes and show dataset-specific panels and legends.

layout(matrix(c(1,1,1,1,2,2,2,2,3,4), 5, 2, byrow = TRUE))
par(mar = c(5, 5, 1, 1), xpd = TRUE, cex.axis = 1)

xlab <- paste("PC1 ", round(pcoa$values$Eigenvalues[1]), "%", sep = "")
ylab <- paste("PC2 ", round(pcoa$values$Eigenvalues[2]), "%", sep = "")

# Panel: 2019_2020
plot(pcoa$vectors[,1], pcoa$vectors[,2],
     pch = pch, col = "white", bg = "white",
     xlab = xlab, ylab = ylab, main = "2019_2020")
points(pcoa$vectors[DS_2019_2020,1], pcoa$vectors[DS_2019_2020,2],
       pch = pch[DS_2019_2020], col = "black",
       bg = color_yday[yday[DS_2019_2020]],
       cex = 1.0 + as.numeric(salinity[DS_2019_2020]) / 10)

# Panel: 2013
plot(pcoa$vectors[,1], pcoa$vectors[,2],
     pch = pch, col = "white", bg = "white",
     xlab = xlab, ylab = ylab, main = "2013")
points(pcoa$vectors[DS_2013,1], pcoa$vectors[DS_2013,2],
       pch = pch[DS_2013], col = "black",
       bg = color_yday[yday[DS_2013]],
       cex = 1.0 + as.numeric(salinity[DS_2013]) / 10)

# Legend: day of year colour bar
plot(1:365, rep(1, 365), col = color_yday, pch = "|", cex = 3, axes = FALSE, ylim = c(0.9, 1.3))
axis(1, at = c(1, 182, 365), labels = c("12", "182", "365"), cex = 3)
text(182, 1.2, "Day of year", cex = 2)

# Legend: salinity size scale
plot(c(0, 10, 20), rep(1, 3),
     col = "black", pch = 1, cex = c(1.6, 2.4, 3.2),
     xlim = c(-5, 25), ylim = c(0.9, 1.3), axes = FALSE, ylab = "", xlab = "")
axis(1, at = c(0, 10, 20), labels = c("2", "18", "34"), cex = 3)
text(10, 1.2, "Salinity", cex = 2)

Interpretation: the PCoA shows that community composition is structured by both salinity and season.

---

← Previous · Overview · Next →