PGLS-models

R code of PGLS models testing AMR against life-history and ecological variables

Publication

The following code has been developed in collaboration with Jose Valdebenito. The methodology underlying the use of the script has been detailed in the article "Proximity to humans is associated with antimicrobial-resistant enteric pathogens in wild bird microbiomes" published in Current Biology.

Files Summary

R Packages

phytools
caper
geiger
knitr

Load tree phylogeny

t <- read.tree("/path/to/directory/tree_submit")
t <- drop.tip(t, tip="Larus_californicus")
plotTree(t, ftype="i")

Load data

df <- read.csv("/path/to/directory/Table S7_New ecol.csv")
df$Weights... <- NULL
df$Common.name <- NULL
colnames(df) <- c("species", "number", "lin", "amr", "prop","b.mass","clutch_size", "msys", "migra", "habit", "diet", "social")

df$b.mass <- log(df$b.mass)
df$habit <- relevel(factor(df$habit), ref="terrestrial")
df$migra <- relevel(factor(df$migra), ref="no")
df$diet <- relevel(factor(df$diet), ref="omnivorous")
df$social <- as.factor(df$social)
df$msys <- as.factor(df$msys)

df <- df[match(t$tip.label, df$species),]

set.seed(4)

Then we prepare for running the phylogenetically corrected least squares (PGLS). It will be four sets of models, one for each measure of AMR:

a.Shannon diversity of AMR b.Shannon diversity of Clonal complexes (CC) c.The proportion of AMR genes present per host species, excluding genes against beta-lactam drugs

comp1 <- comparative.data(data=df, phy=t, species, vcv = T, vcv.dim = 3)

a.Shannon diversity of AMR Run all models, one after the other [(AMR parameters and Ecological variables (N=30)]

pgls1 <- pgls(amr ~ social, comp1, lambda = "ML")
pgls2 <- pgls(amr ~ msys, comp1, lambda = "ML")
pgls3 <- pgls(amr ~ migra, comp1, lambda = "ML")
pgls4 <- pgls(amr ~ habit, comp1, lambda = "ML")
pgls5 <- pgls(amr ~ diet+b.mass, comp1, lambda = "ML")
pgls6 <- pgls(amr ~ b.mass, comp1, lambda = "ML")
pgls7 <- pgls(amr ~ clutch_size, comp1, lambda = "ML")

pgls.res <- round(rbind(
  summary(pgls1)$coefficients,
  summary(pgls2)$coefficients,
  summary(pgls3)$coefficients,
  summary(pgls4)$coefficients,
  summary(pgls5)$coefficients,
  summary(pgls6)$coefficients,
  summary(pgls7)$coefficients),3)

pgls.res

We continue with remaining three set of models: b.Shannon diversity of Clonal complexes (CC) c.The proportion of AMR genes present per host species, excluding genes against beta-lactam drugs

pgls1.1 <- pgls(lin ~ social, comp1, lambda = "ML")
pgls2.1 <- pgls(lin ~ msys, comp1, lambda = "ML")
pgls3.1 <- pgls(lin ~ migra, comp1, lambda = "ML")
pgls4.1 <- pgls(lin ~ habit, comp1, lambda = "ML")
pgls5.1 <- pgls(lin ~ diet+b.mass, comp1, lambda = "ML")
pgls6.1 <- pgls(lin ~ b.mass, comp1, lambda = "ML")
pgls7.1 <- pgls(lin ~ clutch_size, comp1, lambda = "ML")

pgls.res1 <- round(rbind(
  summary(pgls1.1)$coefficients,
  summary(pgls2.1)$coefficients,
  summary(pgls3.1)$coefficients,
  summary(pgls4.1)$coefficients,
  summary(pgls5.1)$coefficients,
  summary(pgls6.1)$coefficients,
  summary(pgls7.1)$coefficients),3)

pgls1.2 <- pgls(prop ~ social, comp1, lambda = "ML")
pgls2.2 <- pgls(prop ~ msys, comp1, lambda = "ML")
pgls3.2 <- pgls(prop ~ migra, comp1, lambda = "ML")
pgls4.2 <- pgls(prop ~ habit, comp1, lambda = "ML")
pgls5.2 <- pgls(prop ~ diet+b.mass, comp1, lambda = "ML")
pgls6.2 <- pgls(prop ~ b.mass, comp1, lambda = "ML")
pgls7.2 <- pgls(prop ~ clutch_size, comp1, lambda = "ML")

pgls.res2 <- round(rbind(
  summary(pgls1.2)$coefficients,
  summary(pgls2.2)$coefficients,
  summary(pgls3.2)$coefficients,
  summary(pgls4.2)$coefficients,
  summary(pgls5.2)$coefficients,
  summary(pgls6.2)$coefficients,
  summary(pgls7.2)$coefficients),3)

pgls.res1

Lineage diversity

pgls.res2

Proportion of AMR

Proceed to do the conduct the part of the analysis involving the proximity to humans estimates AMR parameters and Proximity to urbanisation (N=38) amr = AMR diversity lin = Lineage diversity prop = Proportion of resistant isolates (excluding β-lactam genes)

tt <- read.tree("/path/to/directory/tree_submit_add_branch")
#loading three 38 tips. this one has sister branches of repeated species of length <0.00001.
tt <- drop.tip(tt, tip="Larus_californicus")
tt

pgd <- read.csv("/path/to/directory/Table S7_New prox.csv", header = T)
pgd$Weights... <- NULL
colnames(pgd)[1:8] <- c("species", "name", "country", "prox", "number", "lin", "amr", "prop")
comp <- comparative.data(tt, pgd, species, vcv=T,vcv.dim = 3)

pg1 <- pgls(amr ~ prox, comp, lambda = "ML")
pg2 <- pgls(lin ~ prox, comp, lambda = "ML")
pg3 <- pgls(plogis(prop) ~ prox, comp, lambda = "ML")
summary(pg1)

summary(pg2)

summary(pg3)

AMR parameters and proximity to humans, subset excluding genomes with n<2 (N=20)

pgd0 <- pgd[pgd$number!=1,]
comp0 <- comparative.data(tt, pgd0, species, vcv=T,vcv.dim = 3)
pg10 <- pgls(amr ~ prox, comp0, lambda = "ML")
pg20 <- pgls(lin ~ prox, comp0, lambda = "ML")
pg30 <- pgls(plogis(prop) ~ prox, comp0, lambda = "ML")
summary(pg10)

summary(pg20)

summary(pg30)

How to cite

Mourkas E, Valdebenito JO, Marsh H, Hitchings MD, Cooper KK, Parker CT, Székely T, Johansson H, Ellström P, Pascoe B, Waldenström J, Sheppard SK (2024) Proximity to humans is associated with antimicrobial resistant enteric pathogens in wild bird microbiomes. Current Biology. doi: https://doi.org/10.1016/j.cub.2024.07.059

evangelosmourkas / PGLS-models

readme