1
2
Figure S1: Fits of seven models to rarefied phylogenetic gain values from Gibbons et al. 2013.
1
3
4
5
6
7
Figure S2: Sensitivity of neutral model predictions to the probability of an immigrant replacing
open sites in the local community (m). Local-global overlap in phylogenetic diversity is plotted
as a function of the ratio of community size and global richness (N/S).
8
9
10
11
2
12
13
14
15
Table S1: Data sources used for phylogenetic overlap and global richness estimation. Local
samples are extremely deeply sequenced samples all derived from a single study (Caporaso et al.
2011). The global datasets are broadly spatially distributed surveys at the continental to global
scale, and were obtained from the Qiime and Earth Microbiome Project databases.
Local samples
Human gut
Human tongue
Human skin
Ocean
Lake
Sample ID
Amz30adlt.418837
Amz1teenF.418404
USygt15.F.418688
M31Tong
F11Tong
M31Plmr
M11Plmr
F11Plmr
seqs_I1.631407
NP3
NP5
LMEpi24M
# sequences
3.8M
3.8M
3.5M
2.1M
1.8M
0.75M
0.47M
0.20M
10.7M
1.2M
1.4M
2.3M
Source
QIIME DB study 805
QIIME DB study 805
QIIME DB study 805
EMP study 722
EMP study 722
EMP study 722
EMP study 722
EMP study 722
ERP001778
EMP study 722
EMP study 722
EMP study 722
1.5M
EMP study 722
CC1
SV1
CL3
Description
Amazon adult
Amazon teen
U.S. teen
Male #31
Female #11
Male #31
Male #11
Female#11
English Channel
Newport Pier sample 3
Newport Pier sample 5
Lake Mendota, WI,
US
Sparkling Lake, WI,
US
Cedar Creek
Sevilleta LTER
Calhoun LTSE
1.5M
1M
1.1M
EMP study 722
EMP study 722
EMP study 722
Source
QIIME DB study 850
QIIME DB study 968
QIIME DB study 969
QIIME DB study 968
QIIME DB study 969
ICOMM
EMP study 945
EMP study 1041
QIIME DB study 1552
EMP study 1242
EMP study 1288
EMP study 1627
QIIME DB study 619
EMP study 808
QIIME DB study 1792
QIIME DB study 103
QIIME DB study 213
EMP study 1702
QIIME DB study 397
# sequences
1,094M
1.8M
2.9M
1.9M
1.6M
9.1M
59M
4.5M
0.26M
45M
67M
1.4M
0.37M
2.5M
3.8M
0.15M
0.12M
1.9M
0.038M
% of sequences used
~1%
100%
100%
100%
100%
100%
1.7%
22%
100%
1.5%
100%
100%
100%
44%
26%
100%
100%
52%
100%
Source
# sequences
% of sequences used
QIIME DB study 850
QIIME DB study 968,
969
QIIME DB study 968,
969
ICOMM
EMP study 945
EMP study 632
1,094M
4.7M
~1%
100%
3.5M
100%
9.1M
59M
2.8M
100%
9%
100%
SLEpi20M
Soil
Global datasets
Human Gut
Human Tongue
Human Skin
Ocean
Lake
Soil
Sets used to estimate
global OTU richness (S)
Human gut
Human Tongue
Human Skin
Ocean
Lake
Soil
16
17
3
18
Table S2: Justifications of estimated environmental sample sizes used in Fig. 2.
Environment
Human gut
Human tongue
Human skin
Ocean
Lake
Soil
Estimated typical
sample size
1 x 1011 cells
1 x 107 cells
1 x 104 cells
2 x 108 cells
2 x 108 cells
1 x 108 cells
Justification
Franks et al. 1998 Appl Environ Microbiol
Swab covers approximately 5% of mouth
~4 cm2 swab
2000 ml sample x 1e5 cells/ml
200 ml sample x 1e6 cells/ml
one gram of soil
19
20
4
21
Code for neutral model of local-global overlap of phylogenetic diversity
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
rm(list=ls())
69
70
71
72
73
74
75
####### iterate across variety of global diversities
# set global diversities to consider
Ss=c(10,50,100,500,1000,5000,10000,50000,100000) #Note: 5e4 and 1e5 can take
a long time to run
library(ape)
library(picante)
# function for running neutral community simulations
untbJones<function(start,prob=0,D=1,gens=150,keep=FALSE,meta=rep(1/length(start),length
(start))){
# start: vector of starting abundances (length of start should be richness of
metacommunity);
# sum of start is local population size
# prob: probability of immigrant vs. internal reproduction
# D: number of deaths per generation
# gens: number of generations to simulate
# keep: if true, keep all time points
# meta: vector representing probabilities of immigration from metacommunity;
related to metacommunity SAD
out=matrix(0,gens,length(start))
out[1,]=start
N=sum(start)
#local population size
S=length(start) #metacommunity diversity
for(i in 2:gens){
trans=rep(0,S)
deaths=table(sample(rep(1:S,times=out[(i-1),]),D,replace=TRUE))
trans[as.numeric(names(deaths))]=-deaths
Nbirths=sum(runif(D)>prob)
births=table(sample(1:S,Nbirths,replace=TRUE,prob=(out[(i1),]+trans)/(N-D)))
trans[as.numeric(names(births))]=trans[as.numeric(names(births))]+births
imms=table(sample(1:S,D-Nbirths,replace=TRUE,prob=meta))
trans[as.numeric(names(imms))]=trans[as.numeric(names(imms))]+imms
out[i,]=out[(i-1),]+trans
}
colnames(out)=1:S
if(keep){
return(out)
}else{
return(out[gens,])
}
}
# set replicates per global diversity and matrix to store results
reps=10
5
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
collectSims=matrix(NA,reps,length(Ss))
collectPhyloSor=collectSims
# local population size
N=1000
# number of generations
gens=1000
# iterate across global diversities
for(i in 1:length(Ss)){
# generate global SAD (log normal)
meanSAD=-13
# average mean of SAD estimated from observations
sdSAD=1.43
# average sd of SAD estimated from observations
lnDraw=rlnorm(Ss[i],meanSAD,sdSAD)
SAD=lnDraw/sum(lnDraw)
#global species abundance distribution
# repeat process reps times
for(j in 1:reps){
# create starting local community
sam=table(sample(1:Ss[i],N,replace=TRUE))
start=rep(0,Ss[i])
start[as.numeric(names(sam))]=sam
names(start)=1:Ss[i]
tree=rtree(Ss[i],tip.label=1:Ss[i],rooted=TRUE)
cur=untbJones(start=start,prob=0.5,D=1,gens=gens,keep=FALSE,meta=SAD)
collectSims[j,i]=sum(cur>0)/Ss[i]
collectPhyloSor[j,i]=phylosor(rbind(1,(cur>0)*1),tree)
}
}
#Summarize replicate model runs
UNiterMeans=colMeans(collectSims)
UNiterSDs=apply(collectSims,2,sd)
PSiterMeans=colMeans(collectPhyloSor)
PSiterSDs=apply(collectPhyloSor,2,sd)
PSiter=rbind(PSiterMeans,PSiterSDs)
colnames(PSiter)=Ss
colnames(storePhyloSor)=Ss
# Fit model to predictions
NSs=log10(N/Ss)
# Likelihood function for fitting statistical model to simulation results
fitSig<-function(p,x,y){
yhat=(x-p[1])^p[3]/(p[2]+(x-p[1])^p[3])
-sum(dnorm(y,yhat,exp(p[4]),log=TRUE))
}
# Estimate statistical model parameters
CompOverlapFit=optim(c(2,0,2,1),fitSig,x=NSs,y=UNiterMeans,control=list(maxit=1000000))
dev.new()
6
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
# Plot statistical model predictions
plot(log10(N/Ss),UNiterMeans,pch=16,cex=1,xlab="log10(N/S)",ylab="Composition
al overlap",ylim=c(0,1),xlim=c(-2,8))
arrows(log10(N/Ss),UNiterMeans+2*UNiterSDs,log10(N/Ss),UNiterMeans2*UNiterSDs,code=0)
fit=CompOverlapFit
lines(xp,(xp-fit$par[1])^fit$par[3]/(fit$par[2]+(xpfit$par[1])^fit$par[3]),lwd=2)
# Repeat statistical model fits and plotting for phylogenetic overlap
PhyloSorFit=optim(c(2,0,2,1),fitSig,x=NSs,y=PSiterMeans,control=list(maxit=1000000))
dev.new()
plot(log10(N/Ss),PSiterMeans,pch=16,cex=1,xlab="log10(N/S)",ylab="Phylogeneti
c overlap",ylim=c(0,1),xlim=c(-2,8))
arrows(log10(N/Ss),PSiterMeans+2*PSiterSDs,log10(N/Ss),PSiterMeans2*PSiterSDs,code=0)
fit=PhyloSorFit
lines(xp,(xp-fit$par[1])^fit$par[3]/(fit$par[2]+(xpfit$par[1])^fit$par[3]),lwd=2)
7