# 231002 # Hamamoto et al. all codes # R version 4.3.1 (2023-06-16) ####contig number comfirmation at each step#### #Input data d1 <- read.delim("stats_input.tsv", sep=" ", stringsAsFactors=FALSE, header=TRUE, na.strings="", check.names=FALSE) summary(d1[,2]) sum(d1[,2]) sd(d1[,2]) #Mean 72249, SD 7788.601 #After cutadapt d1 <- read.delim("stats_cutadapt.tsv", sep=" ", stringsAsFactors=FALSE, header=TRUE, na.strings="", check.names=FALSE) summary(d1[,2]) sum(d1[,2]) sd(d1[,2]) #Mean 71044, SD 7660.764 #After dada2 library(data.table) d1 <- fread("stats_dada2.tsv", sep=" ", stringsAsFactors=FALSE, header=TRUE, na.strings="", check.names=FALSE) d1 <- data.frame(d1) rn <- d1$id rownames(d1) <- rn d1 <- d1[,-1] #removing coluimns that contains zero contigs d2 <- d1[,colSums(d1)!=0] #counting column numbers（=number of ASV） ncol(d2)#86842 #calculating sum, mean and SD of contigs of each sample dada <- rowSums(d2) sum(dada)#4647794 mean(dada)#34428.1 sd(dada)#5671.343 #After filtering d1 <- fread("stats_filtered.tsv", sep=" ", stringsAsFactors=FALSE, header=TRUE, na.strings="", check.names=FALSE) d1 <- data.frame(d1) rn <- d1$id rownames(d1) <- rn d1 <- d1[,-1] #removing coluimns that contains zero contigs d2 <- d1[,colSums(d1)!=0] #counting column numbers（=number of ASV） ncol(d2)#83585 #calculating sum, mean and SD of contigs of each sample filtering <- rowSums(d2) sum(filtering)#4171086 mean(filtering)#30896.93 sd(filtering)#5957.975 #After rarefaction d1 <- fread("stats_rarefacted.tsv", sep=" ", stringsAsFactors=FALSE, header=TRUE, na.strings="", check.names=FALSE) d1 <- data.frame(d1) rn <- d1$id rownames(d1) <- rn d1 <- d1[,-1] #removing coluimns that contains zero contigs d2 <- d1[,colSums(d1)!=0] #counting column numbers（=number of ASV） ncol(d2)#83490 #calculating sum, mean and SD of contigs of each sample rarefacted <- rowSums(d2) sum(rarefacted)#3859931 mean(rarefacted)#28592.08 sd(rarefacted)#4497.227 #After copy numner correction d1 <- fread("stats_norm.tsv", sep=" ", stringsAsFactors=FALSE, header=FALSE, na.strings="", check.names=FALSE) d1 <- data.frame(d1) rn <- d1[,1] rownames(d1) <- rn d1 <- d1[,-1] #removing coluimns that contains zero contigs d2 <- d1[,colSums(d1)!=0] #counting column numbers（=number of ASV） ncol(d2)#83138 sum(d2)#3417550 #calculating sum, mean and SD of contigs of each sample correction <- rowSums(d2) mean(correction)#25315.19 sd(correction)#4090.389 #final contigs in each subset # read data d1 <- read.csv("level-7.csv", fileEncoding="UTF-8-BOM", stringsAsFactors=FALSE, header=T, row.names=1) #read metadata metadata<- read.csv("metadata.csv") row.names(metadata) <- metadata[,1] d1c<-cbind(metadata,d1)#結合 # Counting the ocntig number of each substrate type readn <- rowsum(d1c[5:ncol(d1c)],d1c$coral_or_sand_or_seagrass, reorder = FALSE) sum(readn[1,])#sand -> 819714.5 sum(readn[2,])#seagrass -> 1031764 sum(readn[3,])#coral -> 1566071 #### The most abundant prokaryote phylum in each substrate type#### # remove all variables rm(list=ls()) # read data d1 <- read.csv("level-2.csv", fileEncoding="UTF-8-BOM", stringsAsFactors=FALSE, header=T, row.names=1) # read matadata metadata　<-　read.csv("metadata.csv") # combine data d1c <- cbind(metadata, d1) # calcurate sum of each prokaryote phylum in each substrate type and swap row and columns d1c2<-rowsum(d1c[5:ncol(d1c)],d1c$substrate, reorder = FALSE) d2<-as.data.frame(t(d1c2)) # show top five phyla and its relative percentage #coral reefs library(dplyr) cs<-select(.data = d2,"coral") head(arrange(cs/sum(cs)*100,-cs$coral),n=5) #d__Bacteria.p__Proteobacteria 33.48867 #d__Bacteria.p__Bacteroidota 15.65878 #d__Bacteria.p__Cyanobacteria 13.96443 #d__Bacteria.p__Actinobacteriota 10.982722 #d__Bacteria.p__Verrucomicrobiota 5.482178 #sandy bottom snd<-select(.data = d2,"sand") head(arrange(snd/sum(snd)*100,-snd$sand),n=5) #d__Bacteria.p__Proteobacteria 33.21630 #d__Bacteria.p__Actinobacteriota 15.34424 #d__Bacteria.p__Bacteroidota 11.52903 #d__Bacteria.p__Acidobacteriota 8.162781 #d__Bacteria.p__Chloroflexi 5.300401 #seagrass meadow sgs<-select(.data = d2,"seagrass") head(arrange(sgs/sum(sgs)*100,-sgs$seagrass),n=5) #d__Bacteria.p__Proteobacteria 37.02067 #d__Bacteria.p__Actinobacteriota 13.87329 #d__Bacteria.p__Bacteroidota 12.49272 #d__Bacteria.p__Acidobacteriota 6.382322 #d__Bacteria.p__Desulfobacterota 5.874890 #Table 1 #### Barplot #### # remove all variables rm(list=ls()) # import data d1 <- read.csv("level-2.csv", fileEncoding="UTF-8-BOM", stringsAsFactors=FALSE, header=T, row.names=1) # library packages library(dplyr) library(reshape2) library(ggplot2) ## making phyla with less than 30000 contigs in "others" others <- colnames(d1[, colSums(d1)< 30000]) Others <- rowSums(d1[, colSums(d1)< 30000]) d1br <- d1[, !colnames(d1) %in% c(others)] taxa_order <- d1br %>% colSums() %>% sort(decreasing = T) %>% data.frame() %>% rownames() d1br <- d1br %>% dplyr::select(all_of(taxa_order)) d1br <- cbind(d1br, Others) # add "others" ## convert config numners into relative percentage d2b <- sweep(d1br, 1, 100 / rowSums(d1), "*") # import metadata metadata<-read.csv("metadata.csv") # combining the metadata and data then transform dc <- cbind(metadata$substrate,metadata$X, d2b) colnames(dc)[c(1,2)] <- c("substrate","SampleID") df_bar <- melt(dc, id.vars = c("SampleID","substrate"), variable.name = "OTU") df_bar$substrate <- factor(df_bar$substrate, levels = rev(unique(df_bar$substrate))) # plot bar_phylum <- "NULL" bar_phylum <- df_bar %>% group_by(substrate) %>% ggplot(aes(x = SampleID, y = value, fill = OTU)) + geom_bar(stat = "identity", position = "fill", col = "black") + labs(x = "Samples", y = "Relative abundance", fill = "OTU")+ facet_wrap(~substrate, scales = "free")　+ theme( axis.title.x = element_text(size = 18, face = "bold"), axis.text.x = element_text(size = 5, angle = 90, vjust = 0.5), axis.title.y = element_text(size = 18, face = "bold"), axis.text.y = element_text(size = 5, hjust = 0.5), legend.key = element_blank(), legend.text = element_text(size = 5) )+ scale_fill_manual(values = c(palette.colors(n = 12, palette = "Paired"),"#00ffff","lightgray"), guide = guide_legend(byrow=T, nrow=21, title.vjust=1, label.hjust=0)) + coord_flip() bar_phylum #Figure 2 #### Non-hierarchical analysis (nMDS) #### # remove all variables rm(list=ls()) # import data d1 <- read.csv("level-7.csv", fileEncoding="UTF-8-BOM", stringsAsFactors=FALSE, header=T, row.names=1) # calculate distance library(vegan) dis.bc <- vegdist(d1, method="bray") # Bray-Curtis dissimilarity #calculate silhouette values require(cluster) asilmk <- numeric(nrow(d1)) for (i in 2:(nrow(d1)-1)){ asilmk[i] <- pam(dis.bc, k=i, diss=TRUE)$silinfo$avg.width } asilmk which.max(asilmk) # three clusters plot(asilmk, type="b") # plot details of silhouette values and clustering resbc.pam <- pam(dis.bc, k=which.max(asilmk), diss=TRUE) summary(resbc.pam) plot(resbc.pam, cex.axis=0.1) # write a csv file of clustering metadata<-read.csv("metadata.csv") clust1 <- data.frame(resbc.pam$clustering,metadata$substrate) write.csv(clust1, "selclust1.csv") # plot nMDS grp <- as.numeric(resbc.pam$clustering) res1.mds <- metaMDS(dis.bc, trace=0) data.scores <- as.data.frame(res1.mds$points) data.scores$site <- rownames(data.scores) data.scores$grp <- grp head(data.scores) # converting the groupings into labels data.scores["grp"] <- lapply(data.scores["grp"], gsub, pattern="1", replacement = "A") data.scores["grp"] <- lapply(data.scores["grp"], gsub, pattern="2", replacement = "B") data.scores["grp"] <- lapply(data.scores["grp"], gsub, pattern="3", replacement = "C") # preparation for plotting grp.a <- data.scores[data.scores$grp == "A", ][chull(data.scores[data.scores$grp == "A", c("MDS1", "MDS2")]), ] grp.b <- data.scores[data.scores$grp == "B", ][chull(data.scores[data.scores$grp == "B", c("MDS1", "MDS2")]), ] grp.c <- data.scores[data.scores$grp == "C", ][chull(data.scores[data.scores$grp == "C", c("MDS1", "MDS2")]), ] hull.data <- rbind(grp.a, grp.b,grp.c) library(ggplot2) library(ggrepel) ggplot() + geom_polygon(data=hull.data,aes(x=MDS1,y=MDS2,fill=grp,group=grp), alpha=0.30) + geom_text(data=data.scores,aes(x=MDS1,y=MDS2,label=site), alpha=0.5, size=2) + geom_point(data=data.scores,aes(x=MDS1,y=MDS2,shape=grp,colour=grp),size=2) + scale_colour_manual(values=c("A" = "red", "B" = "blue", "C" = "green")) + geom_text_repel()+ coord_equal() + theme_bw() + theme( axis.title.x = element_text(size=10), axis.title.y = element_text(size=10), axis.text.x = element_text(size=10), axis.text.y = element_text(size=10), ) # Figure 3 #### Alpha diversity Index#### # remove all variables rm(list=ls()) # import data d1 <- read.csv("level-7.csv", fileEncoding="UTF-8-BOM", stringsAsFactors=FALSE, header=T, row.names=1) # import metadata metadata <- read.csv("metadata.csv") # calculate diversity index library(vegan) library(ggsignif) Div1 <- diversity(d1)#Shannon-Wiener index #Div2 <- diversity(d1c[1:ncol(d1)], "simpson") #Simpson's 1-D # combine data Div1b<-cbind(Div1,metadata) #Div2b<-cbind(Div2,factors) # test data normality tapply(Div1b$Div1,Div1b$substrate,shapiro.test) # sandy bottom and seagrass meadow doesn't distribute normaly #　perform Kruskal-Wallis rank sum test result <- kruskal.test(Div1b$Div1~Div1b$substrate) result # conduct post-hoc test # Bonferroni method pairwise.wilcox.test(Div1b$Div1, Div1b$substrate, p.adj="bonferroni", exact=F) # drow boxplot library(ggplot2) library(ggsignif) bplot <- ggplot(Div1b, aes(y=Div1,x=substrate)) + geom_boxplot(width=0.6) + ggtitle("Prokaryotic alpha-diversity")+ ylab("Shannon index") + xlab("Substratum type")+ theme(plot.title = element_text(hjust = 0.5))+ geom_signif(comparisons = list(c("Sandy bottom", "Seagrass meadow")), map_signif_level=TRUE, y_position = 5.7) bplot #Figure 4 #### PERMANOVA#### # remove all variables rm(list=ls()) # import data d1 <- read.csv("level-7.csv", fileEncoding="UTF-8-BOM", stringsAsFactors=FALSE, header=T, row.names=1) # import metadata metadata<-read.csv("metadata.csv") rownames(metadata) <- metadata[,1] d1c<-cbind(metadata,d1) # define groups X1 <- metadata$Sites X2 <- metadata$coral_or_noncoral X3 <- metadata$substrate # conduct PERMANOVA adonis2(d1~X3,method = "bray",p.adjust.m = "bonferroni",perm = 5000) # posthoc test library(devtools) library(pairwiseAdonis) pair.mod2 <- pairwise.adonis(d1c[5:ncol(d1c)],factors=d1c$substrate,sim.method = "bray",p.adjust.m = "bonferroni",perm = 999) pair.mod2 #Table 2 #### ANCOM #### # remove all variables rm(list=ls()) # import data d1 <- read.csv("level-4.csv", fileEncoding="UTF-8-BOM", stringsAsFactors=FALSE, header=T, row.names=1) # import libraries library(phyloseq) library(ANCOMBC) library(dplyr) library(ggplot2) library(gridExtra) # extract top 10% orders d2 <- sort(colSums(dr),decreasing = TRUE) otu_name <- head(names(d2), n=as.numeric(ncol(dr))/10) top_tenpercent <- dplyr::select(d1, any_of(otu_name)) # import metadata metadata　<-　read.csv("metadata.csv") dc　<-　cbind(metadata,top_tenpercent) # prepare otu_table and sample_data objects row.names(metadata)<-row.names(d1) sd <-sample_data(metadata) do <- otu_table(top_tenpercent, F) # create a phyloseq file dp<-phyloseq(otu_table(do),sample_data(sd)) # Perform the ANCOMBC test out = ancombc2(data = dp, tax_level = NULL, pseudo_sens = FALSE, fix_formula = "substrate", p_adj_method = "holm", s0_perc = 0.05, prv_cut = 0.10, lib_cut = 0, group = "substrate", alpha = 0.05, verbose = TRUE, global = TRUE, pairwise = TRUE) res_prim = out$res res_global = out$res_global res_pair = out$res_pair row.names(res_pair)<-res_prim$taxon res_dunn = out$res_dunn res_trend = out$res_trend pseudo_sens = out$pseudo_sens_tab # remove orders that assigned to non-scientific name on SILVA library(stringr) res_pair <- filter(res_pair, str_detect(res_pair$taxon, "WS2|WPS.2|Sva0485|SAR324_clade.Marine_group_B.|RCP2.54|NB1.j|MBNT15|LCP.89|FCPU426|.__.__.__", negate = T)) res_pair <- filter(res_pair, str_detect(res_pair$taxon, "o__Pla4_lineage|o__Milano.WF1B.44|o__S085|o__pItb.vmat.80|o__AT.s3.28|o__uncultured|o__JG30.KF.CM66|o__IMCC26256|Subgroup|o__EPR3968.O8a.Bc78|o__HOC36|o__KD4.96|o__SAR202_clade|c__Alphaproteobacteria.__|o__PAUC43f_marine_benthic_group|o__Blfdi19|o__B2M28" , negate = T)) # remove orders having no significant p-values among comparisons any_true <- res_pair$diff_substratesand == TRUE| res_pair$diff_substrateseagrass == TRUE| res_pair$diff_substrateseagrass_substratesand == TRUE res_pair <- filter(res_pair,any_true) # preparation for drawing a boxplot of W（log fold change devided by standard error） coral_vs_sand <- data.frame(res_pair$W_substratesand, res_pair$q_substratesand, row.names = res_pair$taxon) %>% rename("W_statistics" = res_pair.W_substratesand, "q_value" = res_pair.q_substratesand)%>% mutate(comparison = "coral_vs_sand", star=ifelse(q_value<.001, "***", ifelse(q_value<.01, "**", ifelse(q_value<.05, "*", "")))) coral_vs_seagrass <- data.frame(res_pair$W_substrateseagrass, res_pair$q_substrateseagrass, row.names = res_pair$taxon) %>% rename("W_statistics" = res_pair.W_substrateseagrass, "q_value" = res_pair.q_substrateseagrass)%>% mutate(comparison = "coral_vs_seagrass", star=ifelse(q_value<.001, "***", ifelse(q_value<.01, "**", ifelse(q_value<.05, "*", "")))) sand_vs_seagrass <- data.frame(res_pair$W_substrateseagrass_substratesand, res_pair$q_substrateseagrass_substratesand, row.names = res_pair$taxon) %>% rename("W_statistics" = res_pair.W_substrateseagrass_substratesand, "q_value" = res_pair.q_substrateseagrass_substratesand)%>% mutate(comparison = "sand_vs_seagrass", star=ifelse(q_value<.001, "***", ifelse(q_value<.01, "**", ifelse(q_value<.05, "*", "")))) # Combine dataframes into a dataframe, and set factors df_ancom <- rbind(coral_vs_sand,coral_vs_seagrass,sand_vs_seagrass) %>% mutate(order = rep(rownames(coral_vs_sand),3)) df_ancom$comparison <- factor(df_ancom$comparison,levels = c("coral_vs_sand","coral_vs_seagrass","sand_vs_seagrass")) df_ancom$order <- factor(df_ancom$order,levels = rev(rownames(coral_vs_sand))) # sort with realtive abundance nm1 <- colnames(top_tenpercent) nm2 <- rownames(coral_vs_sand) nm3 <- intersect(nm1,nm2) # cutting strings with "__" to extract order names #tmp <- data.frame(str_split(df_ancom$order, pattern = "__", n=5)) #tmp <- t(tmp) #order_name <- as.vector(tmp[,4]) # plot bar_ancom <- "NULL" bar_ancom <- df_ancom %>% ggplot(aes(x = order, y = W_statistics)) + geom_col(aes(x = order, y = W_statistics, fill = ifelse(W_statistics > 0,"Red","Blue"))) + labs(x = "Order", y = "W_statistics")+ facet_grid(.~comparison)　+ geom_text(aes(y=W_statistics+4*sign(W_statistics), label=star), vjust=.7, color="black", position=position_dodge(width = .5))+ theme(axis.text.y = element_text(size = 6, face = "italic"), strip.text.x = element_text(size = 10), strip.text.y = element_text(size = 14), legend.text = element_text(size = 12), plot.title = element_text(hjust = 0.5, size = 15), panel.grid.major = element_blank(), axis.ticks = element_blank(), legend.position = "none") + coord_flip() bar_ancom #Figure 5_left # preparation for plotting the abundance abun_name <- unique(df_ancom$order) abun_order <- top_tenpercent %>% dplyr::select(any_of(abun_name)) %>% t() %>% rowSums() %>% data.frame() %>% mutate(order = abun_name) colnames(abun_order) <- c("Abundance","Order") #　plot bar_abun <- abun_order %>% ggplot(aes(x = Order, y = Abundance)) + geom_bar(stat = "identity") + labs(x = "Abundance")+ coord_flip() + theme(axis.title.x = element_text(size = 10), axis.title.y = element_blank(), axis.text.y = element_blank()) bar_abun #Figure 5_right #### SECOM　#### # remove all variables rm(list=ls()) # set a function (from original article) get_upper_tri = function(cormat){ cormat[lower.tri(cormat)] = NA diag(cormat) = NA return(cormat) } # import data d1 <- read.csv("level-4.csv", fileEncoding="UTF-8-BOM", stringsAsFactors=FALSE, header=T, row.names=1) # library packages library(phyloseq) library(ANCOMBC) library(dplyr) library(tidyr) library(tibble) library(ggplot2) library(gridExtra) library(stringr) # import metadata metadata　<-　read.csv("metadata.csv") row.names(metadata)<-row.names(d1) # extract top 20 orders for visualization d2 <- sort(colSums(d1),decreasing = TRUE) head(d2, n=20) otu_name <- head(names(d2),n=20) # remove orders with ambiguous assignment otu_name <- otu_name[-c(14,16,17,18)] # what removed are #[1] "d__Bacteria.p__Proteobacteria.c__Gammaproteobacteria.__" #[2] "d__Bacteria.p__Proteobacteria.c__Gammaproteobacteria.o__HOC36" #[3] "d__Bacteria.p__Chloroflexi.c__KD4.96.o__KD4.96" #[4] "d__Bacteria.p__NB1.j.c__NB1.j.o__NB1.j" top_ten <- select(d1, any_of(otu_name)) # rename for plotting ct <- regexpr(".o__", otu_name) order_name <- substr(otu_name, ct, 100) colnames(top_ten) <- order_name d3 <- cbind(metadata,top_ten) # prepare individual dataset of coral reef, sady bottom and seagrass meadow samples # coral reef coral_tmp1 <- filter(d3, d3$substrate == "coral") coral_tmp2 <- coral_tmp1[,5:ncol(coral_tmp1)] coral <- coral_tmp2[rowSums(coral_tmp2)!=0,colSums(coral_tmp2)!=0] # create otu_table and sample_data objects sd_coral <-sample_data(coral_tmp1[,1:4]) do_coral <- otu_table(coral, F) # create a phyloseq object dp_coral <- phyloseq(otu_table(do_coral),sample_data(sd_coral)) # secom_dist set.seed(123) res_sd_coral<- secom_dist( data = list(dp_coral), tax_level = NULL, pseudo = 0, prv_cut = 0.5, lib_cut = 1000, corr_cut = 0.5, wins_quant = c(0.05, 0.95), R = 1000, thresh_hard = 0, max_p = 0.005, n_cl = 1 ) # remove samples with low Co-occurrence values (=having many blanks) corr_sd_coral <- res_sd_coral$dcorr_fl cooccur_sd_coral <- res_sd_coral$mat_cooccur overlap = 10 corr_sd_coral[cooccur_sd_coral < overlap] = 0 # p-value sd_coral_p.val = data.frame(get_upper_tri(res_sd_coral$dcorr_p)) %>% rownames_to_column("var1") %>% pivot_longer(cols = -var1, names_to = "var2", values_to = "p_value")%>% filter(!is.na(p_value)) #swap rows with columns df_dist_coral = data.frame(get_upper_tri(corr_sd_coral)) %>% rownames_to_column("var1") %>% pivot_longer(cols = -var1, names_to = "var2", values_to = "value") %>% filter(!is.na(value)) %>% mutate(var2 = gsub("\\...", " - ", var2), value = round(value, 2), metric = "Distance", p_value = sd_coral_p.val$p_value, divide = "Coral")%>% mutate(star = ifelse(p_value<.001, "***", ifelse(p_value<.01, "**", ifelse(p_value<.05, "*", "")))) # secom_linear set.seed(123) res_sl_coral <- secom_linear( data = list(dp_coral), tax_level = NULL, pseudo = 0, prv_cut = 0.5, lib_cut = 1000, corr_cut = 0.5, wins_quant = c(0.05, 0.95), method = "pearson", soft = FALSE, thresh_len = 100, n_cv = 10, thresh_hard = 0, max_p = 0.005, n_cl = 1 ) # removing ones have low Co-occurrence (= having many lacking values) corr_sl_coral <- res_sl_coral$corr_fl cooccur_sl_coral <- res_sl_coral$mat_cooccur overlap = 10 corr_sl_coral[cooccur_sl_coral < overlap] = 0 # extract the p-values sl_coral_p.val = data.frame(get_upper_tri(res_sl_coral$corr_p)) %>% rownames_to_column("var1") %>% pivot_longer(cols = -var1, names_to = "var2", values_to = "p_value")%>% filter(!is.na(p_value)) # pivot the table to longer form df_linear_coral = data.frame(get_upper_tri(corr_sl_coral)) %>% rownames_to_column("var1") %>% pivot_longer(cols = -var1, names_to = "var2", values_to = "value") %>% filter(!is.na(value)) %>% mutate(var2 = gsub("\\...", " - ", var2), value = round(value, 2), metric = "Pearson", p_value = sl_coral_p.val$p_value, divide = "Coral")%>% mutate(star = ifelse(p_value<.001, "***", ifelse(p_value<.01, "**", ifelse(p_value<.05, "*", "")))) #sand sand_tmp1 <- filter(d3, d3$substrate == "sand") sand_tmp2 <- sand_tmp1[,5:ncol(sand_tmp1)] sand <- sand_tmp2[rowSums(sand_tmp2)!=0,colSums(sand_tmp2)!=0] # create otu_table and sample_data objects sd_sand <-sample_data(sand_tmp1[,1:4]) do_sand <- otu_table(sand, F) # create a phyloseq object dp_sand <- phyloseq(otu_table(do_sand),sample_data(sd_sand)) # secom_dist set.seed(123) res_sd_sand<- secom_dist( data = list(dp_sand), tax_level = NULL, pseudo = 0, prv_cut = 0.5, lib_cut = 1000, corr_cut = 0.5, wins_quant = c(0.05, 0.95), R = 1000, thresh_hard = 0, max_p = 0.005, n_cl = 1 ) # remove samples with low Co-occurrence values (=having many blanks) corr_sd_sand <- res_sd_sand$dcorr_fl cooccur_sd_sand <- res_sd_sand$mat_cooccur overlap = 10 corr_sd_sand[cooccur_sd_sand < overlap] = 0 #extract p-values sd_sand_p.val = data.frame(get_upper_tri(res_sd_sand$dcorr_p)) %>% rownames_to_column("var1") %>% pivot_longer(cols = -var1, names_to = "var2", values_to = "p_value")%>% filter(!is.na(p_value)) # pivot the table to longer form df_dist_sand = data.frame(get_upper_tri(corr_sd_sand)) %>% rownames_to_column("var1") %>% pivot_longer(cols = -var1, names_to = "var2", values_to = "value") %>% filter(!is.na(value)) %>% mutate(var2 = gsub("\\...", " - ", var2), value = round(value, 2), metric = "Distance", p_value = sd_sand_p.val$p_value, divide = "Sand")%>% mutate(star = ifelse(p_value<.001, "***", ifelse(p_value<.01, "**", ifelse(p_value<.05, "*", "")))) # secom_linear set.seed(123) res_sl_sand <- secom_linear( data = list(dp_sand), tax_level = NULL, pseudo = 0, prv_cut = 0.5, lib_cut = 1000, corr_cut = 0.5, wins_quant = c(0.05, 0.95), method = "pearson", soft = FALSE, thresh_len = 100, n_cv = 10, thresh_hard = 0, max_p = 0.005, n_cl = 1 ) #remove samples with low Co-occurrence values (=having many blanks) corr_sl_sand <- res_sl_sand$corr_fl cooccur_sl_sand <- res_sl_sand$mat_cooccur overlap = 10 corr_sl_sand[cooccur_sl_sand < overlap] = 0 # extract p-values sl_sand_p.val = data.frame(get_upper_tri(res_sl_sand$corr_p)) %>% rownames_to_column("var1") %>% pivot_longer(cols = -var1, names_to = "var2", values_to = "p_value")%>% filter(!is.na(p_value)) #pivot the table to longer form df_linear_sand = data.frame(get_upper_tri(corr_sl_sand)) %>% rownames_to_column("var1") %>% pivot_longer(cols = -var1, names_to = "var2", values_to = "value") %>% filter(!is.na(value)) %>% mutate(var2 = gsub("\\...", " - ", var2), value = round(value, 2), metric = "Pearson", p_value = sl_sand_p.val$p_value, divide = "Sand")%>% mutate(star = ifelse(p_value<.001, "***", ifelse(p_value<.01, "**", ifelse(p_value<.05, "*", "")))) #seagrass seagrass_tmp1 <- filter(d3, d3$substrate == "seagrass") seagrass_tmp2 <- seagrass_tmp1[,5:ncol(seagrass_tmp1)] seagrass <- seagrass_tmp2[rowSums(seagrass_tmp2)!=0,colSums(seagrass_tmp2)!=0] # create otu_table and sample_data objects row.names(metadata)<-row.names(d1) sd_seagrass <-sample_data(seagrass_tmp1[,1:4]) do_seagrass <- otu_table(seagrass, F) # create a phyloseq object dp_seagrass<-phyloseq(otu_table(do_seagrass),sample_data(sd_seagrass)) # secom_dist set.seed(123) res_sd_seagrass<- secom_dist( data = list(dp_seagrass), tax_level = NULL, pseudo = 0, prv_cut = 0.5, lib_cut = 1000, corr_cut = 0.5, wins_quant = c(0.05, 0.95), R = 1000, thresh_hard = 0, max_p = 0.005, n_cl = 1 ) #remove samples with low Co-occurrence values (=having many blanks) corr_sd_seagrass <- res_sd_seagrass$dcorr_fl cooccur_sd_seagrass <- res_sd_seagrass$mat_cooccur overlap = 10 corr_sd_seagrass[cooccur_sd_seagrass < overlap] = 0 # extract p-values sd_seagrass_p.val = data.frame(get_upper_tri(res_sd_seagrass$dcorr_p)) %>% rownames_to_column("var1") %>% pivot_longer(cols = -var1, names_to = "var2", values_to = "p_value")%>% filter(!is.na(p_value)) #pivot the table to longer form df_dist_seagrass = data.frame(get_upper_tri(corr_sd_seagrass)) %>% rownames_to_column("var1") %>% pivot_longer(cols = -var1, names_to = "var2", values_to = "value") %>% filter(!is.na(value)) %>% mutate(var2 = gsub("\\...", " - ", var2), value = round(value, 2), metric = "Distance", p_value = sd_seagrass_p.val$p_value, divide = "Seagrass")%>% mutate(star = ifelse(p_value<.001, "***", ifelse(p_value<.01, "**", ifelse(p_value<.05, "*", "")))) # secom_linear set.seed(123) res_sl_seagrass <- secom_linear( data = list(dp_seagrass), tax_level = NULL, pseudo = 0, prv_cut = 0.5, lib_cut = 1000, corr_cut = 0.5, wins_quant = c(0.05, 0.95), method = "pearson", soft = FALSE, thresh_len = 100, n_cv = 10, thresh_hard = 0, max_p = 0.005, n_cl = 1 ) #remove samples with low Co-occurrence values (=having many blanks) corr_sl_seagrass <- res_sl_seagrass$corr_fl cooccur_sl_seagrass <- res_sl_seagrass$mat_cooccur overlap = 10 corr_sl_seagrass[cooccur_sl_seagrass < overlap] = 0 # extract p-values sl_seagrass_p.val = data.frame(get_upper_tri(res_sl_seagrass$corr_p)) %>% rownames_to_column("var1") %>% pivot_longer(cols = -var1, names_to = "var2", values_to = "p_value")%>% filter(!is.na(p_value)) #pivot the table to longer form df_linear_seagrass = data.frame(get_upper_tri(corr_sl_seagrass)) %>% rownames_to_column("var1") %>% pivot_longer(cols = -var1, names_to = "var2", values_to = "value") %>% filter(!is.na(value)) %>% mutate(var2 = gsub("\\...", " - ", var2), value = round(value, 2), metric = "Pearson", p_value = sl_seagrass_p.val$p_value, divide = "Seagrass")%>% mutate(star = ifelse(p_value<.001, "***", ifelse(p_value<.01, "**", ifelse(p_value<.05, "*", "")))) # conbine all tables df_secom = df_linear_coral %>% bind_rows( df_dist_coral ) %>% bind_rows( df_linear_sand ) %>% bind_rows( df_dist_sand ) %>% bind_rows( df_linear_seagrass ) %>% bind_rows( df_dist_seagrass ) df_secom$var1 = str_sub(df_secom$var1,5) df_secom$var2 = str_sub(df_secom$var2,5) phylum_level = sort(union(df_secom$var1, df_secom$var2)) df_secom$var1 = factor(df_secom$var1, levels = phylum_level) df_secom$var2 = factor(df_secom$var2, levels = phylum_level) df_secom$metric = factor(df_secom$metric, levels = unique(df=secom$divide)) # plot heat_secom = df_secom %>% ggplot(aes(var2, var1, fill = value)) + geom_tile(color = "black") + scale_fill_gradient2(low = "blue", high = "red", mid = "white", na.value = "grey", midpoint = 0, limit = c(-1,1), space = "Lab", name = NULL) + scale_x_discrete(drop = TRUE) + scale_y_discrete(drop = TRUE) + facet_grid(divide~metric) + geom_text(aes(var2, var1, label = value,), color = "black", size = 2) + #geom_text(aes(var2, var1, label = star), color = "black", size = 3) + labs(x = NULL, y = NULL, title = "Correlations") + theme_bw() + theme(axis.text.x = element_text(angle = 45, vjust = 1, size = 6, hjust = 1, face = "italic"), axis.text.y = element_text(size = 6, face = "italic"), strip.text.x = element_text(size = 14), strip.text.y = element_text(size = 14), legend.text = element_text(size = 12), plot.title = element_text(hjust = 0.5, size = 15), panel.grid.major = element_blank(), axis.ticks = element_blank(), legend.position = "none") + coord_fixed() heat_secom ggsave("~/Desktop/paper publication/03'. microbial community in coral rich and absent sediments/SECOM/SECOM_combined_order.pdf", heat_secom, width = 8, height = 15, dpi = 300) #### Enzyme~diveristy indices　#### # remove all variables rm(list=ls()) # import data d1 <- read.csv("en_level_4.csv", fileEncoding="UTF-8-BOM", stringsAsFactors=FALSE, header=T, row.names=1, check.names = FALSE) # import metadata metadata<- read.csv("metadata.csv") # calcurate the diversity index library(vegan) library(ggsignif) Div1 <- diversity(d1)#Shannon-Weaver index #Div2 <- diversity(d1, "simpson") #Simpson's 1-D # combine data and metadata Div1b<-cbind(Div1,metadata) #Div2b<-cbind(Div2,metadata) # test normality tapply(Div1b$Div1,Div1b$substrate,shapiro.test) # conduct anova test followed by Tukey's HSD test result <- TukeyHSD(aov(data = Div1b,Div1~substrate)) result # Tukey multiple comparisons of means #95% family-wise confidence level #Fit: aov(formula = Div1 ~ substrate, data = Div1b) #$substrate # diff lwr upr p adj #sand-coral -0.054152863 -0.06826383 -0.040041894 0.0000000 #seagrass-coral -0.060817469 -0.07435998 -0.047274962 0.0000000 #seagrass-sand -0.006664606 -0.02202037 0.008691154 0.5600666 # plot a boxplot library(ggplot2) library(ggsignif) bplot <- ggplot(Div1b, aes(y=Div1,x=substrate)) + geom_boxplot(width=0.6) + ggtitle("Enzymic alpha-diversity")+ ylab("Shannon index") + xlab("Substratum type")+ theme(plot.title = element_text(hjust = 0.5))+ geom_signif(comparisons = list(c("Coral reef", "Sandy bottom"),c("Coral reef", "Seagrass meadow")), map_signif_level=TRUE, y_position = c(6.75,6.77)) bplot #ggsave("~/Desktop/paper publication/03'. microbial community in coral rich and absent sediments/enzymes/barplot_type.pdf", bplot, width = 5, height = 5, dpi = 300) # test normality of individual sites tapply(Div1b$Div1,Div1b$Sites,shapiro.test) #conduct anova test followed by TukeyHSD result2 <- TukeyHSD(aov(data = Div1b, Div1~Sites)) result2 # Tukey multiple comparisons of means # 95% family-wise confidence level # #Fit: aov(formula = Div1 ~ Sites, data = Div1b) # #$Sites #diff lwr upr p adj #Ky-Kn 0.012895193 -0.0092432732 0.0350336592 0.5873552 #Mn-Kn 0.088009104 0.0658706382 0.1101475706 0.0000000 #Od-Kn 0.067190975 0.0450525089 0.0893294413 0.0000000 #Sk-Kn 0.080803763 0.0586652965 0.1029422289 0.0000000 #Sn-Kn 0.057201864 0.0332895876 0.0811141406 0.0000000 #Ur-Kn 0.022805764 0.0006672983 0.0449442307 0.0389376 #Mn-Ky 0.075113911 0.0529754452 0.0972523776 0.0000000 #Od-Ky 0.054295782 0.0321573159 0.0764342483 0.0000000 #Sk-Ky 0.067908570 0.0457701035 0.0900470359 0.0000000 #Sn-Ky 0.044306671 0.0203943946 0.0682189476 0.0000032 #Ur-Ky 0.009910571 -0.0122278947 0.0320490377 0.8312014 #Od-Mn -0.020818129 -0.0429565955 0.0013203369 0.0799280 #Sk-Mn -0.007205342 -0.0293438079 0.0149331245 0.9585067 #Sn-Mn -0.030807240 -0.0547195168 -0.0068949638 0.0033198 #Ur-Mn -0.065203340 -0.0873418061 -0.0430648737 0.0000000 #Sk-Od 0.013612788 -0.0085256786 0.0357512538 0.5224390 #Sn-Od -0.009989111 -0.0339013875 0.0139231654 0.8722990 #Ur-Od -0.044385211 -0.0665236768 -0.0222467445 0.0000004 #Sn-Sk -0.023601899 -0.0475141751 0.0003103779 0.0555377 #Ur-Sk -0.057997998 -0.0801364644 -0.0358595320 0.0000000 #Ur-Sn -0.034396100 -0.0583083761 -0.0104838231 0.0006307 # plot bplot2 <- ggplot(Div1b, aes(y=Div1,x=Sites)) + geom_boxplot(width=0.6) + ggtitle("Enzymic alpha-diversity")+ ylab("Shannon index") + xlab("Sites")+ theme(plot.title = element_text(hjust = 0.5)) bplot2 #Figure S2 ##### Enzyme~ANCOM #### # remove all variables rm(list=ls()) # import data d1 <- read.csv("en_level_4.csv", fileEncoding="UTF-8-BOM", stringsAsFactors=FALSE, header=T, row.names=1, check.names = FALSE) # library packages library(phyloseq) library(ANCOMBC) library(stringr) library(dplyr) library(ggplot2) library(gridExtra) #　extract top 5% of all exzymes d2 <- sort(colSums(d1),decreasing = TRUE) head(names(d2), n=as.numeric(ncol(d1))/20) otu_name <- head(names(d2), n=as.numeric(ncol(d1))/20) top_fivepercent <- select(d1, any_of(otu_name)) #summary(colSums(dr)) #dr <- dr[,colSums(dr)>1271246] # import metadata factors　<-　read.csv("factors.csv") # create otu_table and sample_data objects row.names(factors)<-row.names(d1) sd <-sample_data(factors) do <- otu_table(top_fivepercent, F) # create a phyloseq object dp<-phyloseq(otu_table(do),sample_data(sd)) # perform ANCOMBC test out = ancombc2(data = dp, tax_level = NULL, pseudo_sens = FALSE, fix_formula = "substrate", p_adj_method = "holm", s0_perc = 0.05, prv_cut = 0.10, lib_cut = 0, group = "substrate", alpha = 0.05, verbose = TRUE, global = TRUE, pairwise = TRUE) res_prim = out$res res_global = out$res_global res_pair = out$res_pair row.names(res_pair)<-res_prim$taxon res_dunn = out$res_dunn res_trend = out$res_trend pseudo_sens = out$pseudo_sens_tab # remove orders having no significant p-values among comparisons any_true <- res_pair$diff_substratesand == TRUE| res_pair$diff_substrateseagrass == TRUE| res_pair$diff_substrateseagrass_substratesand == TRUE res_pair <- filter(res_pair,any_true) # preparation for drawing a boxplot of W（log fold change devided by standard error） coral_vs_sand <- data.frame(res_pair$W_substratesand, res_pair$q_substratesand, row.names = res_pair$taxon) %>% rename("W_statistics" = res_pair.W_substratesand, "q_value" = res_pair.q_substratesand)%>% mutate(comparison = "coral_vs_sand", star=ifelse(q_value<.001, "***", ifelse(q_value<.01, "**", ifelse(q_value<.05, "*", "")))) coral_vs_seagrass <- data.frame(res_pair$W_substrateseagrass, res_pair$q_substrateseagrass, row.names = res_pair$taxon) %>% rename("W_statistics" = res_pair.W_substrateseagrass, "q_value" = res_pair.q_substrateseagrass)%>% mutate(comparison = "coral_vs_seagrass", star=ifelse(q_value<.001, "***", ifelse(q_value<.01, "**", ifelse(q_value<.05, "*", "")))) sand_vs_seagrass <- data.frame(res_pair$W_substrateseagrass_substratesand, res_pair$q_substrateseagrass_substratesand, row.names = res_pair$taxon) %>% rename("W_statistics" = res_pair.W_substrateseagrass_substratesand, "q_value" = res_pair.q_substrateseagrass_substratesand)%>% mutate(comparison = "sand_vs_seagrass", star=ifelse(q_value<.001, "***", ifelse(q_value<.01, "**", ifelse(q_value<.05, "*", "")))) # Combine dataframes into a dataframe, and set factors df_ancom <- rbind(coral_vs_sand,coral_vs_seagrass,sand_vs_seagrass) %>% mutate(Function = rep(rownames(coral_vs_sand),3)) df_ancom$comparison <- factor(df_ancom$comparison,levels = c("coral_vs_sand","coral_vs_seagrass","sand_vs_seagrass")) df_ancom$Function <- factor(df_ancom$Function,levels = rev(rownames(coral_vs_sand))) # plot bar_ancom <- "NULL" bar_ancom <- df_ancom %>% ggplot(aes(x = Function, y = W_statistics)) + geom_col(aes(x = Function, y = W_statistics, fill = ifelse(W_statistics > 0,"Red","Blue"))) + labs(x = "Function", y = "W_statistics")+ facet_grid(.~comparison)　+ geom_text(aes(y=W_statistics+4*sign(W_statistics), label=star), vjust=.7, color="black", position=position_dodge(width = .5))+ theme(axis.text.y = element_text(size = 6, face = "italic"), strip.text.x = element_text(size = 10), strip.text.y = element_text(size = 14), legend.text = element_text(size = 12), plot.title = element_text(hjust = 0.5, size = 15), panel.grid.major = element_blank(), axis.ticks = element_blank(), legend.position = "none") + coord_flip() bar_ancom # Figure 7 left # preparation for plotting the abundance abun_name <- unique(df_ancom$Function) abun_enzyme <- top_fivepercent %>% dplyr::select(any_of(abun_name)) %>% t() %>% rowSums() %>% data.frame() %>% mutate(order = abun_name) colnames(abun_enzyme) <- c("Abundance","Function") #　plot bar_abun <- abun_enzyme %>% ggplot(aes(x = Function, y = Abundance)) + geom_bar(stat = "identity") + labs(x = "Abundance")+ coord_flip() + theme(axis.title.x = element_text(size = 10), axis.title.y = element_blank(), axis.text.y = element_blank()) bar_abun # Figure 7 right