Load packages

library("ampvis")

Load data

data(DNAext_1.0)

Subset to the relevant dataset

pcr <- subset_samples(V13, Exp.PCR == "YES") %>%
  rarefy_even_depth(sample.size = 25000, rngseed = 712) %>%
  filter_taxa(function(x) max(x) >= 10, TRUE)

Overall differences between samples using PCA

PCA with square root transformed OTU abundances.

pca <- amp_ordinate(data = pcr,
                    plot.color = "Add.Label", 
                    plot.point.size = 5,
                    plot.group = "chull",
                    plot.group.label = "Add.Label",                  
                    output = "complete",
                    envfit.factor = "Add.Label",
                    envfit.show = F,
                    plot.theme = "clean")

Plot the PCA. It looks like there might be some significant grouping.

pca$plot + theme(legend.position = "none")

The model reports a p-value of 0.001, hence there is an effect of different PCR settings.

pca$eff.model
## 
## ***FACTORS:
## 
## Centroids:
##                        PC1     PC2
## Add.Label1 ng      -2.2568  1.6732
## Add.Label25 cycles -2.1403  0.6196
## Add.Label35 cycles  0.0822  0.0188
## Add.Label50 ng      2.8735  1.1218
## Add.Label52 C       0.0454 -4.4422
## Add.Label58 C       2.3664  1.2207
## Add.LabelStandard  -0.9431 -0.2056
## 
## Goodness of fit:
##               r2 Pr(>r)    
## Add.Label 0.8995  0.001 ***
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## Permutation: free
## Number of permutations: 999

Overall differences between samples using beta diversity

The Bray-Curtis dissimilarity index is used as an alternative method to test for significant groupings in the dataset.

beta <- amp_test_cluster(data = pcr, group = "Add.Label", method = "bray", plot.color = "Add.Label", plot.label = "Add.Label")

Using adonis we also find a significant effect as the p-value is 0.001.

beta$adonis
## 
## Call:
## adonis(formula = test_formula, data = sample) 
## 
## Permutation: free
## Number of permutations: 999
## 
## Terms added sequentially (first to last)
## 
##           Df SumsOfSqs   MeanSqs F.Model      R2 Pr(>F)    
## Add.Label  6  0.068761 0.0114601   2.217 0.50574  0.001 ***
## Residuals 13  0.067200 0.0051692         0.49426           
## Total     19  0.135960                   1.00000           
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1

Figure 6A: Influence of PCR settings

amp_ordinate(data = pcr, 
             plot.color = "Add.Label", 
             plot.point.size = 2,
             plot.group = "chull",
             plot.theme = "clean"
             ) +
  xlim(-4,4) +
  annotate("text", x = -2, y = 3, label = "1 ng", size = 2) +
  annotate("text", x = -1, y = -0.9, label = "Standard", size = 2) +
  annotate("text", x = -1, y = -1.4, label = "56*' '*degree*C*', 5 ng, 30 cyc'", size = 2, parse = T) +
  annotate("text", x = -3.4, y = 0.5, label = "25 cyc", size = 2) +
  annotate("text", x = 0.8, y = -4, label = "52*' '*degree*C", size = 2, parse = T) +  
  annotate("text", x = 0.9, y = 0.3, label = "35 cyc", size = 2) +
  annotate("text", x = 2.7, y = 0.3, label = "50 ng", size = 2) +  
  annotate("text", x = 2, y = 1.8, label = "58*' '*degree*C", size = 2, parse = T) +     
  theme(legend.position = "none",
        axis.text = element_text(size = 6, color = "black"), 
        text = element_text(size = 8, color = "black")
        )

Save the plot

ggsave("plots/Fig6A.eps", width = 50, height = 50, units = "mm")