In tapj/IBSMicrobiota: Tap, Derrien et al article R source code.

library(knitr)
opts_chunk$set(fig.width=12, fig.height=8, warning=FALSE, message=FALSE, 
echo=FALSE, dev=c("png", "pdf"), fig.cap=TRUE, cache=TRUE)

library(ggplot2)
library(reshape2)
library(RColorBrewer)
library(IBSMicrobiota)
library(dplyr)
library(magrittr)

First we load IBS diet, OTU and clinical data and merge them

data(IBSData)
data(diet)
data(otuseverity)

metadata = IBSData$metadata
idx.v4 = which( metadata$Visit=="V4" & metadata$Sample_type =="Stool" ) 
# we selected only stool sample baseline = visit V4
metadata = metadata[idx.v4,]

otuseverity_abund    = prop.table(as.matrix(IBSData$otu),2)[otuseverity,rownames(metadata)]
tax                  = IBSData$tax

diet_variable = c("kcal","PROT%","FAT%","CARB%","Total FODMAPs")

diet_metadata = 
merge(diet[,diet_variable], metadata, by.x="row.names", by.y= "Patient_ID", all.x=TRUE) 

diet_metadata_otu = 
merge(diet_metadata, t(otuseverity_abund), by.x="Sample_ID", by.y="row.names", all.x=TRUE)

table(diet_metadata_otu$Health)

diet_metadata_otu = diet_metadata_otu[, c("Age","Gender","BMI",diet_variable,"IBS.Severity.Score","SSgroup",rownames(otuseverity_abund))]

Second, we compute pairwise wilcoxon test between diet and health status

diet_SSgroup_melt = na.omit(melt(diet_metadata_otu[, c(diet_variable,"SSgroup")], id.vars="SSgroup"))

levels(diet_SSgroup_melt$variable) = c("energy intake (kcal)", "Protein (%)", "Fat (%)", "Carb. (%)", "Total FODMAPs (g)")

diet_vs_SSgroup_plot = 
ggplot(diet_SSgroup_melt) +  geom_boxplot(aes(x=SSgroup,y=value, fill=SSgroup)) + 
scale_fill_manual("Health status", values=rev(brewer.pal(n=4, name="BrBG"))) +
facet_wrap(~variable, scale="free") + 
ylab("") + xlab("")


pdf("/home/tapju/storage/Dropbox/Danone/IBSMicrobiota/inst/figures/Tap_diet_vs_SSgroup_plot.pdf", w=11, h=5)
diet_vs_SSgroup_plot 
dev.off()

diet_vs_SSgroup_plot 


result_p=NULL
for(i in levels(diet_SSgroup_melt$variable)) {
    idx = which(diet_SSgroup_melt$variable==i)
    p   = pairwise.wilcox.test(diet_SSgroup_melt$value[idx], diet_SSgroup_melt$SSgroup[idx], p.adjust="fdr")$p.value
    p   = as.vector(p)

    result_p = rbind(result_p,p)
}

result_p = as.data.frame(t(na.omit(t(result_p))))

colnames(result_p)  = c("control vs mild","control vs moderate","control vs severe", "mild vs moderate","mild vs severe","moderate vs severe")
rownames(result_p)  = levels(diet_SSgroup_melt$variable)

result_p #non significant expect for Protein % between mild,control vs moderate

third, we computed the coinertia analysis between diet and OTUs (n=90) from the microbial signature for IBS severity. RV coeficient and Monte-carlo test was used to test how diet and those OTUs were related

diet_otu = na.omit(diet_metadata_otu[, c(diet_variable, rownames(otuseverity_abund))])

diet.pca = dudi.pca(diet_otu[,diet_variable], scannf=F, nf=3)
otu.pca  = dudi.pca(log10(diet_otu[,rownames(otuseverity_abund)] + 10^-5), scannf=F, nf=3)

diet_otu_coi = coinertia(diet.pca,otu.pca, scannf=F, nf=3)

randtest(diet_otu_coi) #non_significant

diet_summary = 
diet_metadata_otu[, c(3,4,5,6,7,23)] %>% melt(., id.vars="SSgroup") %>% 
group_by(SSgroup,variable) %>% 
summarize(median=median(round(value,2)), interquartile.inf=round(quantile(value, 0.25),2), interquartile.sup=round(quantile(value, 0.75),2)) %>%
as.data.frame %>% na.omit #%>% dcast(SSgroup~variable) %>% as.data.frame


diet_summary 

#write.csv2(diet_summary, file="inst/tables/TableS2_diet_summary.csv")

diet_summary =
dcast(
data.frame(diet_summary[1:2], v=paste0(diet_summary$median, " (", diet_summary$interquartile.inf, " - ", diet_summary$interquartile.sup,")"))

, SSgroup~variable, value.var="v")


write.csv2(diet_summary, file="inst/tables/TableS2_diet_summary.csv")

Conclusion

Diet was not associated with symptoms severity group except Protein intake which found to be different between control/mild vs moderate. Diet was not associated with microbial signature OTU for IBS severity.