Supplementary Software

commands for data processing in QIIME and R scripts for statistical analysis 

## Sequence data processing using QIIME
split_libraries.py -m qiime_mapping.txt -f seqs.fna -q seqs.qual -l 300 -b 10 -z truncate_only	#demultiplex sequences 
identify_chimeric_seqs.py -i seqs.fna -m usearch61 -o identified_chimeras --threads 1 -r gg_13_8_otus.fasta
java -XX:+UseConcMarkSweepGC -Xmx150G –jar acacia-1.52.b0.jar -c acacia.config	#error correction, see example config file below
pick_otus.py -i acacia_denoised.fasta -s 0.99 -o uclust_picked_otus_99
pick_rep_set.py -i uclust_picked_otus_99/acacia_denoised_otus.txt -f acacia_denoised.fasta
assign_taxonomy.py -i rep_set.fasta -r gg_13_8_otus.fasta -t gg_13_8_taxonomy.txt -m blast -e 0.001 -o blast_taxonomy.txt
make_otu_table.py -i uclust_picked_otus_99/acacia_denoised_otus.txt -o non_normalised_otu_table_99.biom -t blast_taxonomy.txt
filter_otus_from_otu_table.py -i non_normalised_otu_table_99.biom -o non_normalised_OTU_table_99_no10tons.biom -n 10
biom convert -i non_normalised_OTU_table_99_no10tons.biom -o non_normalised_OTU_table_99_no10tons.txt -b --header-key="taxonomy" --output-metadata-id="taxonomy"	#convert biom format OTU table to tab separated format to remove chloroplast and mitochondria OTUs; see below
sed '/c__Chloroplast/d' non_normalised_OTU_table_99_no10tons.txt | sed -i '/f__mitochondria/d'  > non_normalised_OTU_table_99_no10tons_nochlormito.txt
biom convert -i non_normalised_OTU_table_99_no10tons_nochlormito.txt -o non_normalised_OTU_table_99_no10tons_nochlormito.biom --table-type="OTU table" --process-obs-metadata taxonomy	#convert tab separated OTU table to biom format 
biom summarize-table -i non_normalised_OTU_table_99_no10tons_nochlormito.biom -o tablestatistics.txt

## rarefaction-based normalization and 16S copy number adjustment:
single_rarefaction.py -i non_normalised_OTU_table_99_no10tons_nochlormito.biom -o normalised_OTU_table_99_no10tons_nochlormito_1000.biom -d 1000 
biom convert -i normalised_OTU_table_99_no10tons_nochlormito_1000.biom -o normalised_OTU_table_99_no10tons_nochlormito_1000.txt -b --header-key="taxonomy" --output-metadata-id="taxonomy"
copyrighter -i normalised_OTU_table_99_no10tons_nochlormito_1000.txt 

## calculating alpha diversity:
align_seqs.py - rep_set.fasta -o rep_set_alignment/
filter_alignment.py -i rep_set_alignment/rep_set_aligned.fasta -o rep_set_alignment/
make_phylogeny.py -i rep_set_alignment/rep_set_aligned_pfiltered.fasta -o rep_set_tree.tre
multiple_rarefactions.py -i non_normalised_OTU_table_99_no10tons_nochlormito.biom -m 500 -x 19500 -s 500 -o rare_tables
alpha_diversity.py -i rare_tables/ -msimpson,observed_species,PD_whole_tree,chao1 -o alpha_diversity_out -t rep_set_ tree.tree
collate_alpha.py -i alpha_diversity_out -o alpha_diversity_collated


## ====== example acacia config file:

ANY_DIFF_SIGNIFICANT_FOR_TWO_SEQS=TRUE
AVG_QUALITY_CUTOFF=30
ERROR_MODEL=Balzer
FASTA=TRUE
FASTA_LOCATION=seqs.fna
FASTQ=FALSE
FASTQ_LOCATION=null
FILTER_N_BEFORE_POS=300
FLOW_KEY=TCAG
MAXIMUM_MANHATTAN_DISTANCE=13
MAX_RECURSE_DEPTH=2
MAX_STD_DEV_LENGTH=2
MID_FILE=null
MID_OPTION=NO_MID
MIN_FLOW_TRUNCATION=150
MIN_READ_REP_BEFORE_TRUNCATION=0.0
OUTPUT_DIR=acacia_denoised/
OUTPUT_PREFIX=acacia_denoised
QUAL_LOCATION=null
REPRESENTATIVE_SEQUENCE=Mode
SIGNIFICANCE_LEVEL=-9
SPLIT_ON_MID=FALSE
TRIM_TO_LENGTH=300
TRUNCATE_READ_TO_FLOW=
## ======

## R commands for community analysis
# remove samples with less than 1000 sequences in non_normalised_OTU_table_99_no10tons_nochlormito.txt 
# create metadata file for each sample in non_normalised_OTU_table_99_no10tons_nochlormito.txt -> env.csv

# in R:
read.table("env.csv", header=T, row.names=1 ,sep=',') -> env
read.table("non_normalised_OTU_table_99_no10tons_nochlormito.txt", header=T, row.names=1, sep='\t') -> otu.tmp
(otu.tmp[,-409]) -> otu.tmp

# centered log ratio (CLR) transformation:
# from https://github.com/joey711/shiny-phyloseq/blob/master/panels/paneldoc/Transform.md
gm_mean = function(x, na.rm=TRUE){
  exp(sum(log(x[x > 0 & !is.na(x)]), na.rm=na.rm) / length(x))
}
# from https://github.com/joey711/shiny-phyloseq/blob/master/panels/paneldoc/Transform.md
clr = function(x, base=2){
  x <- log((x / gm_mean(x)), base)
  x[!is.finite(x) | is.na(x)] <- 0.0
  return(x)
}

#Run CLR on the OTU count matrix (rows OTUs, cols Samples)
otu <- t(apply(otu.tmp, MARGIN = 2, FUN = clr))
# beta diversity analysis:
library(vegan)

adonis(otu ~ type * soiltype, data=env, method="euclidean")		# runs a PERMANOVA
rda(otu) -> otu.pca	# runs a principal component analysis

plot(otu.pca, type='n')
ordispider(otu.pca, env$type, col="#dddddd")
points(otu.pca, pch=ifelse(env$type=="root",21,24), cex=0.8, bg=env$soiltype)
legend("bottomleft", legend=unique(factor(env$type)), pch=19, col=unique(factor(env$type)))   # PCA figure


#==== community analysis on root and soils separately:
subset(otu, env$type=="root") -> otu.r
subset(env, env$type=="root") -> env.r
subset(otu, env$type=="soil") -> otu.s
subset(env, env$type=="soil") -> env.s 

# community analysis on soil samples only:
adonis(otu.s ~ soiltype, data=env.s, method="euclidean", permu=3000)
rda(otu.s~env.s$soiltype) -> otu.s.rda 		# runs a redundancy analysis (RDA)

plot(otu.s.rda, type='n')
points(otu.s.rda, pch=24, cex=1, bg=env.s$soiltype)
legend("bottomleft",legend=unique(factor(env.s$soiltype)),pch=19,col=unique(factor(env.s$soiltype))) 	# plots the RDA

# community analysis on root samples only:

adonis(otu.r ~ soiltype, data=env.r, method="euclidean", permu=3000)
rda(otu.r) -> otu.r.pca

plot(otu.r.pca, type='n')
ordispider(otu.r.pca,env.r$soiltype,col="#dddddd")
points(otu.r.pca, pch=21, cex=1, bg=env.r$soiltype)
legend("bottomleft",legend=unique(factor(env.r$soiltype)),pch=19,col=unique(factor(env.r$soiltype)))

## relative contributions of soil chemical characteristics and plant phylogeny to community composition:
# extract index of soil chemical composition:
scores(rda(decostand(env.s[,27:43],method="standardize")))$sites[,1] -> X1
scores(rda(decostand(env.s[,27:43],method="standardize")))$sites[,2] -> X2
scores(rda(decostand(env.r[,27:43],method="standardize")))$sites[,1] -> Y1
scores(rda(decostand(env.r[,27:43],method="standardize")))$sites[,2] -> Y2
# extract scores representing host phylogeny:
read.table("rbcL_distmat.txt",header=T,row.names=1,sep='\t') -> host.dm
rda(host.dm)->host.dm.pca  ## 81.6% represented by first two axes (PC1 and PC2)
scores(host.dm.pca)$sites  # copy these scores into env file
adonis(otu.s ~ X1*X2 + rbcL_distmat_pca_PC1*rbcL_distmat_pca_PC2, data=env.s,method="euc", permu=3000) #for soils
adonis(otu.r ~ Y1*Y2 + rbcL_distmat_pca_PC1*rbcL_distmat_pca_PC2, data=env.r,method="euc", permu=3000) #for roots

# split values in host.dm to root and soil only distance matrices, then import into R:
read.table("rbcL_distmat_root.txt",header=T,row.names=1,sep="\t")->host.r.dm
read.table("rbcL_distmat_soil.txt",header=T,row.names=1,sep="\t")->host.s.dm

# create dm for root and soil communities:
vegdist(otu,method="euc")->otu.dm
vegdist(otu.r,method="euc")->otu.r.dm
vegdist(otu.s,method="euc")->otu.s.dm
# Procrustes analysis:
protest(host.r.dm,otu.r.dm,scale=T,permu=3000)
protest(host.s.dm,otu.s.dm,scale=T,permu=3000)
# mantel test:
mantel(host.r.dm,otu.r.dm,method="spearman")	#roots
mantel(host.s.dm,otu.s.dm,method="spearman")	#soils

## Indicator species analysis using labdsv R package
# convert input OTU count matrix to relative abundance using total sum scaling
decostand(otu.tmp, “total”, MARGIN=2) -> relabun
#Filtering for OTUs that have at least >0.5% rel abun in at least one sample
relabun[,apply(relabun,2,max)>0.005] -> relabun
library(labdsv)
relabun[, colSums(abs(relabun)) != 0]->relabun.nozero
subset(relabun.nozero,env$type=="root")->relabun.nozero.r
subset(relabun.nozero,env$type=="soil")->relabun.nozero.s
# run indicator species analysis:
indval(relabun.nozero,env$type)->indval.rootvssoil
gr <- indval.rootvssoil$maxcls
iv <- indval.rootvssoil$indcls
pv <- indval.rootvssoil$pval
fr.r <- apply(relabun.nozero.r>0, 2, sum)
fr.s <- apply(relabun.nozero.s>0, 2, sum)
avg_relabun.r <- colMeans(relabun.nozero.r)
avg_relabun.s <- colMeans(relabun.nozero.s)
indvalsummary <- data.frame(group=gr, indval=iv, pvalue=pv, freq.r=fr.r, freq.s=fr.s, root_avgrelabun=avg_relabun.r, soil_avgrelabun=avg_relabun.s)
indvalsummary <- indvalsummary[order(indvalsummary$group, -indvalsummary$indval),]

# check p value distribution:
subset(indvalsummary,indvalsummary$group==1)->tmp   # root samples only
hist(tmp$pvalue, breaks=20) #if p-value distribution looks good, proceed with P value adjustment
qv <- p.adjust(tmp$pvalue,method="fdr")
indvalsummary.r <-data.frame(tmp,fdr=qv)

## indicator species using Welch’s t test in STAMP
# get list of >0.5% relative abundance OTUs in relabun, name as high_relabun_otus
otu.tmp[which(rownames(otu.tmp) %in% high_relabun_otus$V1),]->otu.tmp
otu_high_relabun_CLR <- t(apply(otu.tmp, MARGIN = 2, FUN = clr))
# output CLR transformed OTU table for input into STAMP
write.table(t(otu_high_relabun_CLR),file="non_normalised_OTUtable_CLR relabun_pointfivepercOTUs_STAMP.txt", sep = "\t", row.names=T,col.names=T)
## indicator species using sparse partial least squares discriminant analysis (sPLSDA) in mixOmics R package
library(mixOmics)
pca(otu, ncomp=10, logratio='none',) -> otu.pca
plot(otu.pca)  # barplot of variance explained
plsda(otu, env$type, ncomp = 3) -> plsda
perf.plsda <- perf(plsda, validation = 'Mfold', folds = 5, progressBar = TRUE, nrepeat = 50)
plot(perf.plsda, overlay = 'measure', sd=TRUE)
plotIndiv(plsda , comp = c(1,2), ind.names = FALSE, ellipse = TRUE, legend = TRUE, title = 'Cooloola Most Diverse, PLSDA comp 1 - 2')
## pick two components only based on perf.plsda 
splsda.tune = tune.splsda(otu, env$type, ncomp = 2, logratio = 'none', test.keepX = c(seq(5,150, 5)), validation = 'Mfold', folds = 5, dist = 'max.dist', nrepeat = 50)
plot(splsda.tune)  # lines indicate that components 1 to 2 are sufficient.
select.keepX = splsda.tune$choice.keepX[1:2]
splsda.res <- splsda(otu, env$type, ncomp = 2, keepX = select.keepX)

plotIndiv(splsda.res, comp = c(1,2), ind.names = FALSE, ellipse = TRUE, legend = TRUE, title = 'cooloola data, sPLSDA comp 1 - 2')

splsda.perf = perf(splsda.res, validation = 'Mfold', folds = 5, progressBar = TRUE, nrepeat = 50, cpus = 3)
splsda.perf$error.rate
plot(splsda.perf)

head(selectVar(splsda.res, comp = 1)$value) 
plotLoadings(splsda.res, comp = 1, method = 'mean', contrib = 'max')

selectVar(splsda.res, comp = 1)$value

# get total sum scaled relative abundance values and append to sPLSDA output:
subset(relabun,env$type=="root") -> relabun.r
subset(relabun,env$type=="soil") -> relabun.s
subset(env,env$type=="root")->env.r
subset(env,env$type=="soil")->env.s

avg_relabun.r <- colMeans(relabun.r)[selectVar(splsda.res, comp = 1)$name]
avg_relabun.s <- colMeans(relabun.s)[selectVar(splsda.res, comp = 1)$name]

avg_relabun_lycopod <- colMeans((relabun.r)[env.r$clade=="lycopod",])[selectVar(splsda.res, comp = 1)$name]
avg_relabun_fern <- colMeans((relabun.r)[env.r$clade=="fern",])[selectVar(splsda.res, comp = 1)$name]
avg_relabun_gymnosperm <- colMeans((relabun.r)[env.r$clade=="gymnosperm",])[selectVar(splsda.res, comp = 1)$name]
avg_relabun_magnoliid <- colMeans((relabun.r)[env.r$class=="basalangio",])[selectVar(splsda.res, comp = 1)$name]
avg_relabun_eudicot <- colMeans((relabun.r)[env.r$class=="eudicot",])[selectVar(splsda.res, comp = 1)$name]
avg_relabun_monocot <- colMeans((relabun.r)[env.r$class=="monocot",])[selectVar(splsda.res, comp = 1)$name]

# to get the relative abundances for each of these indicator OTUs in a single table:
indic_otus <- data.frame(selectVar(splsda.res, comp = 1)$value,avg_relabun.r,avg_relabun.s,avg_relabun_lycopod,avg_relabun_fern,avg_relabun_gymnosperm,avg_relabun_magnoliid,avg_relabun_eudicot,avg_relabun_monocot)
write.table(indic_otus, file = "splsda indicator OTUs.txt", sep='\t')