knitr::opts_chunk$set(message = FALSE, warning = FALSE, comment = NA,
fig.width = 6.25, fig.height = 5)
library(ANCOMBC)
library(tidyverse)
Sparse Estimation of Correlations among Microbiomes (SECOM) (Lin, Eggesbø, and Peddada 2022) is a methodology that aims to detect both linear and nonlinear relationships between a pair of taxa within an ecosystem (e.g., gut) or across ecosystems (e.g., gut and tongue). SECOM corrects both sample-specific and taxon-specific biases and obtains a consistent estimator for the correlation matrix of microbial absolute abundances while maintaining the underlying true sparsity. For more details, please refer to the SECOM paper.
Download package.
if (!requireNamespace("BiocManager", quietly = TRUE))
install.packages("BiocManager")
BiocManager::install("ANCOMBC")
Load the package.
The HITChip Atlas dataset contains genus-level microbiota profiling with HITChip for 1006 western adults with no reported health complications, reported in (Lahti et al. 2014). The dataset is available via the microbiome R package (Lahti et al. 2017) in phyloseq (McMurdie and Holmes 2013) format.
data(atlas1006, package = "microbiome")
tse = mia::makeTreeSummarizedExperimentFromPhyloseq(atlas1006)
# subset to baseline
tse = tse[, tse$time == 0]
# Re-code the bmi group
tse$bmi = recode(tse$bmi_group,
obese = "obese",
severeobese = "obese",
morbidobese = "obese")
# Subset to lean, overweight, and obese subjects
tse = tse[, tse$bmi %in% c("lean", "overweight", "obese")]
# Create the region variable
tse$region = recode(as.character(tse$nationality),
Scandinavia = "NE", UKIE = "NE", SouthEurope = "SE",
CentralEurope = "CE", EasternEurope = "EE",
.missing = "unknown")
# Discard "EE" as it contains only 1 subject
# Discard subjects with missing values of region
tse = tse[, ! tse$region %in% c("EE", "unknown")]
print(tse)
class: TreeSummarizedExperiment
dim: 130 873
metadata(0):
assays(1): counts
rownames(130): Actinomycetaceae Aerococcus ... Xanthomonadaceae
Yersinia et rel.
rowData names(3): Phylum Family Genus
colnames(873): Sample-1 Sample-2 ... Sample-1005 Sample-1006
colData names(12): age sex ... bmi region
reducedDimNames(0):
mainExpName: NULL
altExpNames(0):
rowLinks: NULL
rowTree: NULL
colLinks: NULL
colTree: NULL
set.seed(123)
# Linear relationships
res_linear = secom_linear(data = list(tse), assay_name = "counts",
tax_level = "Phylum", 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 = 20, n_cv = 10,
thresh_hard = 0.3, max_p = 0.005, n_cl = 2)
# Nonlinear relationships
res_dist = secom_dist(data = list(tse), assay_name = "counts",
tax_level = "Phylum", pseudo = 0,
prv_cut = 0.5, lib_cut = 1000, corr_cut = 0.5,
wins_quant = c(0.05, 0.95), R = 1000,
thresh_hard = 0.3, max_p = 0.005, n_cl = 2)
corr_linear = res_linear$corr_th
cooccur_linear = res_linear$mat_cooccur
# Filter by co-occurrence
overlap = 10
corr_linear[cooccur_linear < overlap] = 0
df_linear = data.frame(get_upper_tri(corr_linear)) %>%
rownames_to_column("var1") %>%
pivot_longer(cols = -var1, names_to = "var2", values_to = "value") %>%
filter(!is.na(value)) %>%
mutate(value = round(value, 2))
tax_name = sort(union(df_linear$var1, df_linear$var2))
df_linear$var1 = factor(df_linear$var1, levels = tax_name)
df_linear$var2 = factor(df_linear$var2, levels = tax_name)
heat_linear_th = df_linear %>%
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 = FALSE) +
scale_y_discrete(drop = FALSE) +
geom_text(aes(var2, var1, label = value), color = "black", size = 4) +
labs(x = NULL, y = NULL, title = "Pearson (Thresholding)") +
theme_bw() +
theme(axis.text.x = element_text(angle = 45, vjust = 1, size = 12, hjust = 1,
face = "italic"),
axis.text.y = element_text(size = 12, 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_linear_th
corr_linear = res_linear$corr_fl
cooccur_linear = res_linear$mat_cooccur
# Filter by co-occurrence
overlap = 10
corr_linear[cooccur_linear < overlap] = 0
df_linear = data.frame(get_upper_tri(corr_linear)) %>%
rownames_to_column("var1") %>%
pivot_longer(cols = -var1, names_to = "var2", values_to = "value") %>%
filter(!is.na(value)) %>%
mutate(value = round(value, 2))
tax_name = sort(union(df_linear$var1, df_linear$var2))
df_linear$var1 = factor(df_linear$var1, levels = tax_name)
df_linear$var2 = factor(df_linear$var2, levels = tax_name)
heat_linear_fl = df_linear %>%
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 = FALSE) +
scale_y_discrete(drop = FALSE) +
geom_text(aes(var2, var1, label = value), color = "black", size = 4) +
labs(x = NULL, y = NULL, title = "Pearson (Filtering)") +
theme_bw() +
theme(axis.text.x = element_text(angle = 45, vjust = 1, size = 12, hjust = 1,
face = "italic"),
axis.text.y = element_text(size = 12, 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_linear_fl
corr_dist = res_dist$dcorr_fl
cooccur_dist = res_dist$mat_cooccur
# Filter by co-occurrence
overlap = 10
corr_dist[cooccur_dist < overlap] = 0
df_dist = data.frame(get_upper_tri(corr_dist)) %>%
rownames_to_column("var1") %>%
pivot_longer(cols = -var1, names_to = "var2", values_to = "value") %>%
filter(!is.na(value)) %>%
mutate(value = round(value, 2))
tax_name = sort(union(df_dist$var1, df_dist$var2))
df_dist$var1 = factor(df_dist$var1, levels = tax_name)
df_dist$var2 = factor(df_dist$var2, levels = tax_name)
heat_dist_fl = df_dist %>%
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 = FALSE) +
scale_y_discrete(drop = FALSE) +
geom_text(aes(var2, var1, label = value), color = "black", size = 4) +
labs(x = NULL, y = NULL, title = "Distance (Filtering)") +
theme_bw() +
theme(axis.text.x = element_text(angle = 45, vjust = 1, size = 12, hjust = 1,
face = "italic"),
axis.text.y = element_text(size = 12, 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_dist_fl
To compute correlations whithin and across different ecosystems, one needs to make sure that there are samples in common across these ecosystems.
# Select subjects from "CE" and "NE"
tse1 = tse[, tse$region == "CE"]
tse2 = tse[, tse$region == "NE"]
# Rename samples to ensure there is an overlap of samples between CE and NE
colnames(tse1) = paste0("Sample-", seq_len(ncol(tse1)))
colnames(tse2) = paste0("Sample-", seq_len(ncol(tse2)))
print(tse1)
class: TreeSummarizedExperiment
dim: 130 578
metadata(0):
assays(1): counts
rownames(130): Actinomycetaceae Aerococcus ... Xanthomonadaceae
Yersinia et rel.
rowData names(3): Phylum Family Genus
colnames(578): Sample-1 Sample-2 ... Sample-577 Sample-578
colData names(12): age sex ... bmi region
reducedDimNames(0):
mainExpName: NULL
altExpNames(0):
rowLinks: NULL
rowTree: NULL
colLinks: NULL
colTree: NULL
class: TreeSummarizedExperiment
dim: 130 181
metadata(0):
assays(1): counts
rownames(130): Actinomycetaceae Aerococcus ... Xanthomonadaceae
Yersinia et rel.
rowData names(3): Phylum Family Genus
colnames(181): Sample-1 Sample-2 ... Sample-180 Sample-181
colData names(12): age sex ... bmi region
reducedDimNames(0):
mainExpName: NULL
altExpNames(0):
rowLinks: NULL
rowTree: NULL
colLinks: NULL
colTree: NULL
set.seed(123)
# Linear relationships
res_linear = secom_linear(data = list(CE = tse1, NE = tse2),
assay_name = c("counts", "counts"),
tax_level = c("Phylum", "Phylum"), 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 = 20, n_cv = 10,
thresh_hard = 0.3, max_p = 0.005, n_cl = 2)
# Nonlinear relationships
res_dist = secom_dist(data = list(CE = tse1, NE = tse2),
assay_name = c("counts", "counts"),
tax_level = c("Phylum", "Phylum"), pseudo = 0,
prv_cut = 0.5, lib_cut = 1000, corr_cut = 0.5,
wins_quant = c(0.05, 0.95), R = 1000,
thresh_hard = 0.3, max_p = 0.005, n_cl = 2)
corr_linear = res_linear$corr_th
cooccur_linear = res_linear$mat_cooccur
# Filter by co-occurrence
overlap = 10
corr_linear[cooccur_linear < overlap] = 0
df_linear = data.frame(get_upper_tri(corr_linear)) %>%
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))
tax_name = sort(union(df_linear$var1, df_linear$var2))
df_linear$var1 = factor(df_linear$var1, levels = tax_name)
df_linear$var2 = factor(df_linear$var2, levels = tax_name)
txt_color = ifelse(grepl("CE", tax_name), "#1B9E77", "#D95F02")
heat_linear_th = df_linear %>%
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 = FALSE) +
scale_y_discrete(drop = FALSE) +
geom_text(aes(var2, var1, label = value), color = "black", size = 4) +
labs(x = NULL, y = NULL, title = "Pearson (Thresholding)") +
theme_bw() +
geom_vline(xintercept = 6.5, color = "blue", linetype = "dashed") +
geom_hline(yintercept = 6.5, color = "blue", linetype = "dashed") +
theme(axis.text.x = element_text(angle = 45, vjust = 1, size = 12, hjust = 1,
face = "italic", color = txt_color),
axis.text.y = element_text(size = 12, face = "italic",
color = txt_color),
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_linear_th
corr_linear = res_linear$corr_th
cooccur_linear = res_linear$mat_cooccur
# Filter by co-occurrence
overlap = 10
corr_linear[cooccur_linear < overlap] = 0
df_linear = data.frame(get_upper_tri(corr_linear)) %>%
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))
tax_name = sort(union(df_linear$var1, df_linear$var2))
df_linear$var1 = factor(df_linear$var1, levels = tax_name)
df_linear$var2 = factor(df_linear$var2, levels = tax_name)
txt_color = ifelse(grepl("CE", tax_name), "#1B9E77", "#D95F02")
heat_linear_fl = df_linear %>%
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 = FALSE) +
scale_y_discrete(drop = FALSE) +
geom_text(aes(var2, var1, label = value), color = "black", size = 4) +
labs(x = NULL, y = NULL, title = "Pearson (Filtering)") +
theme_bw() +
geom_vline(xintercept = 6.5, color = "blue", linetype = "dashed") +
geom_hline(yintercept = 6.5, color = "blue", linetype = "dashed") +
theme(axis.text.x = element_text(angle = 45, vjust = 1, size = 12, hjust = 1,
face = "italic", color = txt_color),
axis.text.y = element_text(size = 12, face = "italic",
color = txt_color),
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_linear_fl
corr_dist = res_dist$dcorr_fl
cooccur_dist = res_dist$mat_cooccur
# Filter by co-occurrence
overlap = 10
corr_dist[cooccur_dist < overlap] = 0
df_dist = data.frame(get_upper_tri(corr_dist)) %>%
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))
tax_name = sort(union(df_dist$var1, df_dist$var2))
df_dist$var1 = factor(df_dist$var1, levels = tax_name)
df_dist$var2 = factor(df_dist$var2, levels = tax_name)
txt_color = ifelse(grepl("CE", tax_name), "#1B9E77", "#D95F02")
heat_dist_fl = df_dist %>%
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 = FALSE) +
scale_y_discrete(drop = FALSE) +
geom_text(aes(var2, var1, label = value), color = "black", size = 4) +
labs(x = NULL, y = NULL, title = "Distance (Filtering)") +
theme_bw() +
geom_vline(xintercept = 6.5, color = "blue", linetype = "dashed") +
geom_hline(yintercept = 6.5, color = "blue", linetype = "dashed") +
theme(axis.text.x = element_text(angle = 45, vjust = 1, size = 12, hjust = 1,
face = "italic", color = txt_color),
axis.text.y = element_text(size = 12, face = "italic",
color = txt_color),
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_dist_fl
R version 4.2.3 (2023-03-15)
Platform: x86_64-apple-darwin17.0 (64-bit)
Running under: macOS Big Sur ... 10.16
Matrix products: default
BLAS: /Library/Frameworks/R.framework/Versions/4.2/Resources/lib/libRblas.0.dylib
LAPACK: /Library/Frameworks/R.framework/Versions/4.2/Resources/lib/libRlapack.dylib
locale:
[1] C/en_US.UTF-8/en_US.UTF-8/C/en_US.UTF-8/en_US.UTF-8
attached base packages:
[1] stats graphics grDevices utils datasets methods base
other attached packages:
[1] caret_6.0-94 lattice_0.20-45 doRNG_1.8.6 rngtools_1.5.2
[5] foreach_1.5.2 DT_0.27 phyloseq_1.42.0 lubridate_1.9.2
[9] forcats_1.0.0 stringr_1.5.0 dplyr_1.1.1 purrr_1.0.1
[13] readr_2.1.4 tidyr_1.3.0 tibble_3.2.1 ggplot2_3.4.1
[17] tidyverse_2.0.0 ANCOMBC_2.0.3
loaded via a namespace (and not attached):
[1] utf8_1.2.3 tidyselect_1.2.0
[3] lme4_1.1-32 RSQLite_2.3.0
[5] htmlwidgets_1.6.2 grid_4.2.3
[7] BiocParallel_1.32.6 gmp_0.7-1
[9] pROC_1.18.0 munsell_0.5.0
[11] ScaledMatrix_1.6.0 codetools_0.2-19
[13] future_1.32.0 withr_2.5.0
[15] colorspace_2.1-0 Biobase_2.58.0
[17] energy_1.7-11 highr_0.10
[19] knitr_1.42 rstudioapi_0.14
[21] stats4_4.2.3 SingleCellExperiment_1.20.1
[23] DescTools_0.99.48 listenv_0.9.0
[25] MatrixGenerics_1.10.0 Rdpack_2.4
[27] labeling_0.4.2 emmeans_1.8.5
[29] GenomeInfoDbData_1.2.9 farver_2.1.1
[31] bit64_4.0.5 rhdf5_2.42.0
[33] parallelly_1.35.0 coda_0.19-4
[35] vctrs_0.6.1 treeio_1.22.0
[37] generics_0.1.3 TH.data_1.1-1
[39] ipred_0.9-14 xfun_0.38
[41] timechange_0.2.0 R6_2.5.1
[43] doParallel_1.0.17 GenomeInfoDb_1.34.9
[45] ggbeeswarm_0.7.1 rsvd_1.0.5
[47] bitops_1.0-7 rhdf5filters_1.10.1
[49] cachem_1.0.7 DelayedArray_0.24.0
[51] scales_1.2.1 multcomp_1.4-23
[53] nnet_7.3-18 beeswarm_0.4.0
[55] rootSolve_1.8.2.3 gtable_0.3.3
[57] beachmat_2.14.0 globals_0.16.2
[59] lmom_2.9 sandwich_3.0-2
[61] timeDate_4022.108 rlang_1.1.0
[63] splines_4.2.3 lazyeval_0.2.2
[65] ModelMetrics_1.2.2.2 checkmate_2.1.0
[67] yaml_2.3.7 reshape2_1.4.4
[69] crosstalk_1.2.0 backports_1.4.1
[71] Hmisc_5.0-1 lava_1.7.2.1
[73] tools_4.2.3 ellipsis_0.3.2
[75] decontam_1.18.0 RColorBrewer_1.1-3
[77] jquerylib_0.1.4 biomformat_1.26.0
[79] proxy_0.4-27 BiocGenerics_0.44.0
[81] MultiAssayExperiment_1.24.0 Rcpp_1.0.10
[83] plyr_1.8.8 base64enc_0.1-3
[85] sparseMatrixStats_1.10.0 zlibbioc_1.44.0
[87] RCurl_1.98-1.12 rpart_4.1.19
[89] TreeSummarizedExperiment_2.6.0 viridis_0.6.2
[91] S4Vectors_0.36.2 zoo_1.8-11
[93] SummarizedExperiment_1.28.0 ggrepel_0.9.3
[95] cluster_2.1.4 DECIPHER_2.26.0
[97] magrittr_2.0.3 data.table_1.14.8
[99] lmerTest_3.1-3 mvtnorm_1.1-3
[101] matrixStats_0.63.0 gsl_2.1-8
[103] hms_1.1.3 evaluate_0.20
[105] xtable_1.8-4 readxl_1.4.2
[107] IRanges_2.32.0 gridExtra_2.3
[109] compiler_4.2.3 scater_1.26.1
[111] crayon_1.5.2 minqa_1.2.5
[113] htmltools_0.5.5 mgcv_1.8-42
[115] tzdb_0.3.0 Formula_1.2-5
[117] expm_0.999-7 Exact_3.2
[119] DBI_1.1.3 MASS_7.3-58.3
[121] boot_1.3-28.1 Matrix_1.5-3
[123] ade4_1.7-22 permute_0.9-7
[125] cli_3.6.1 rbibutils_2.2.13
[127] gower_1.0.1 parallel_4.2.3
[129] igraph_1.4.1 GenomicRanges_1.50.2
[131] pkgconfig_2.0.3 numDeriv_2016.8-1.1
[133] foreign_0.8-84 recipes_1.0.5
[135] scuttle_1.8.4 hardhat_1.2.0
[137] vipor_0.4.5 bslib_0.4.2
[139] DirichletMultinomial_1.40.0 multtest_2.54.0
[141] XVector_0.38.0 prodlim_2019.11.13
[143] estimability_1.4.1 mia_1.6.0
[145] CVXR_1.0-11 yulab.utils_0.0.6
[147] digest_0.6.31 vegan_2.6-4
[149] Biostrings_2.66.0 rmarkdown_2.21
[151] cellranger_1.1.0 tidytree_0.4.2
[153] htmlTable_2.4.1 gld_2.6.6
[155] DelayedMatrixStats_1.20.0 nloptr_2.0.3
[157] lifecycle_1.0.3 nlme_3.1-162
[159] jsonlite_1.8.4 Rhdf5lib_1.20.0
[161] BiocNeighbors_1.16.0 viridisLite_0.4.1
[163] fansi_1.0.4 pillar_1.9.0
[165] fastmap_1.1.1 httr_1.4.5
[167] survival_3.5-5 glue_1.6.2
[169] iterators_1.0.14 bit_4.0.5
[171] class_7.3-21 stringi_1.7.12
[173] sass_0.4.5 blob_1.2.4
[175] BiocSingular_1.14.0 memoise_2.0.1
[177] Rmpfr_0.9-1 future.apply_1.10.0
[179] irlba_2.3.5.1 e1071_1.7-13
[181] ape_5.7-1
Lahti, Leo, Jarkko Salojärvi, Anne Salonen, Marten Scheffer, and Willem M De Vos. 2014. “Tipping Elements in the Human Intestinal Ecosystem.” Nature Communications 5 (1): 1–10.
Lahti, Leo, Sudarshan Shetty, T Blake, J Salojarvi, and others. 2017. “Tools for Microbiome Analysis in R.” Version 1: 10013.
Lin, Huang, Merete Eggesbø, and Shyamal Das Peddada. 2022. “Linear and Nonlinear Correlation Estimators Unveil Undescribed Taxa Interactions in Microbiome Data.” Nature Communications 13 (1): 1–16.
McMurdie, Paul J, and Susan Holmes. 2013. “Phyloseq: An R Package for Reproducible Interactive Analysis and Graphics of Microbiome Census Data.” PloS One 8 (4): e61217.