Contents

mia implements tools for microbiome analysis based on the SummarizedExperiment (Morgan et al. 2020), SingleCellExperiment (Amezquita et al. 2020) and TreeSummarizedExperiment (Huang 2021) infrastructure. Data wrangling and analysis are the main scope of this package.

1 Installation

To install mia, install BiocManager first, if it is not installed. Afterwards use the install function from BiocManager.

if (!requireNamespace("BiocManager", quietly = TRUE))
    install.packages("BiocManager")

BiocManager::install("mia")

2 Load mia

library("mia")

3 Loading a TreeSummarizedExperiment object

A few example datasets are available via mia. For this vignette the GlobalPatterns dataset is loaded first.

data(GlobalPatterns, package = "mia")
se <- GlobalPatterns
se
## class: TreeSummarizedExperiment 
## dim: 19216 26 
## metadata(0):
## assays(1): counts
## rownames(19216): 549322 522457 ... 200359 271582
## rowData names(7): Kingdom Phylum ... Genus Species
## colnames(26): CL3 CC1 ... Even2 Even3
## colData names(7): X.SampleID Primer ... SampleType Description
## reducedDimNames(0):
## mainExpName: NULL
## altExpNames(0):
## rowLinks: a LinkDataFrame (19216 rows)
## rowTree: 1 phylo tree(s) (19216 leaves)
## colLinks: NULL
## colTree: NULL

4 Functions for working with microbiome data

One of the main topics for analysing microbiome data is the application of taxonomic data to describe features measured. The interest lies in the connection between individual bacterial species and their relation to each other.

mia does not rely on a specific object type to hold taxonomic data, but uses specific columns in the rowData of a SummarizedExperiment object. taxonomyRanks can be used to construct a character vector of available taxonomic levels. This can be used, for example, for subsetting.

# print the available taxonomic ranks
colnames(rowData(se))
## [1] "Kingdom" "Phylum"  "Class"   "Order"   "Family"  "Genus"   "Species"
taxonomyRanks(se)
## [1] "Kingdom" "Phylum"  "Class"   "Order"   "Family"  "Genus"   "Species"
# subset to taxonomic data only
rowData(se)[,taxonomyRanks(se)]
## DataFrame with 19216 rows and 7 columns
##            Kingdom        Phylum        Class        Order        Family
##        <character>   <character>  <character>  <character>   <character>
## 549322     Archaea Crenarchaeota Thermoprotei           NA            NA
## 522457     Archaea Crenarchaeota Thermoprotei           NA            NA
## 951        Archaea Crenarchaeota Thermoprotei Sulfolobales Sulfolobaceae
## 244423     Archaea Crenarchaeota        Sd-NA           NA            NA
## 586076     Archaea Crenarchaeota        Sd-NA           NA            NA
## ...            ...           ...          ...          ...           ...
## 278222    Bacteria           SR1           NA           NA            NA
## 463590    Bacteria           SR1           NA           NA            NA
## 535321    Bacteria           SR1           NA           NA            NA
## 200359    Bacteria           SR1           NA           NA            NA
## 271582    Bacteria           SR1           NA           NA            NA
##              Genus                Species
##        <character>            <character>
## 549322          NA                     NA
## 522457          NA                     NA
## 951     Sulfolobus Sulfolobusacidocalda..
## 244423          NA                     NA
## 586076          NA                     NA
## ...            ...                    ...
## 278222          NA                     NA
## 463590          NA                     NA
## 535321          NA                     NA
## 200359          NA                     NA
## 271582          NA                     NA

The columns are recognized case insensitive. Additional functions are available to check for validity of taxonomic information or generate labels based on the taxonomic information.

table(taxonomyRankEmpty(se, "Species"))
## 
## FALSE  TRUE 
##  1413 17803
head(getTaxonomyLabels(se))
## [1] "Class:Thermoprotei"               "Class:Thermoprotei_1"            
## [3] "Species:Sulfolobusacidocaldarius" "Class:Sd-NA"                     
## [5] "Class:Sd-NA_1"                    "Class:Sd-NA_2"

For more details see the man page ?taxonomyRanks.

4.1 Merging and agglomeration based on taxonomic information.

Agglomeration of data based on these taxonomic descriptors can be performed using functions implemented in mia. In addition to the aggValue functions provide by TreeSummarizedExperiment agglomerateByRank is available. agglomerateByRank does not require tree data to be present.

agglomerateByRank constructs a factor to guide merging from the available taxonomic information. For more information on merging have a look at the man page via ?mergeRows.

# agglomerate at the Family taxonomic rank
x1 <- agglomerateByRank(se, rank = "Family")
## How many taxa before/after agglomeration?
nrow(se)
## [1] 19216
nrow(x1)
## [1] 603

Tree data can also be shrunk alongside agglomeration, but this is turned of by default.

# with agglomeration of the tree
x2 <- agglomerateByRank(se, rank = "Family",
                        agglomerateTree = TRUE)
nrow(x2) # same number of rows, but
## [1] 603
rowTree(x1) # ... different
## 
## Phylogenetic tree with 19216 tips and 19215 internal nodes.
## 
## Tip labels:
##   549322, 522457, 951, 244423, 586076, 246140, ...
## Node labels:
##   , 0.858.4, 1.000.154, 0.764.3, 0.995.2, 1.000.2, ...
## 
## Rooted; includes branch lengths.
rowTree(x2) # ... tree
## 
## Phylogenetic tree with 496 tips and 260 internal nodes.
## 
## Tip labels:
##   Family:Cenarchaeaceae, Family:SAGMA-X, Family:Nitrososphaeraceae, Family:Sulfolobaceae, Family:Halobacteriaceae, Family:MSBL1, ...
## Node labels:
##   root:ALL, Kingdom:Archaea, Phylum:Crenarchaeota, Class:C2, Class:Sd-NA, Class:Thaumarchaeota, ...
## 
## Rooted; includes branch lengths.

For agglomerateByRank to work, taxonomic data must be present. Even though only one rank is available for the enterotype dataset, agglomeration can be performed effectively de-duplicating entries for the genus level.

data(enterotype, package = "mia")
taxonomyRanks(enterotype)
## [1] "Genus"
agglomerateByRank(enterotype)
## class: TreeSummarizedExperiment 
## dim: 552 280 
## metadata(0):
## assays(1): counts
## rownames(552): -1 Prosthecochloris ... Syntrophococcus Mogibacterium
## rowData names(1): Genus
## colnames(280): AM.AD.1 AM.AD.2 ... TS98_V2 TS99.2_V2
## colData names(9): Enterotype Sample_ID ... Age ClinicalStatus
## reducedDimNames(0):
## mainExpName: NULL
## altExpNames(0):
## rowLinks: NULL
## rowTree: NULL
## colLinks: NULL
## colTree: NULL

To keep data tidy, the agglomerated data can be stored as an alternative experiment in the object of origin. With this synchronized sample subsetting becomes very easy.

altExp(se, "family") <- x2

Keep in mind, that if you set na.rm = TRUE, rows with NA or similar value (defined via the empty.fields argument) will be removed. Depending on these settings different number of rows will be returned.

x1 <- agglomerateByRank(se, rank = "Species", na.rm = TRUE)
altExp(se,"species") <- agglomerateByRank(se, rank = "Species", na.rm = FALSE)
dim(x1)
## [1] 944  26
dim(altExp(se,"species"))
## [1] 2307   26

For convenience the function splitByRanks is available, which agglomerates data on all ranks selected. By default all available ranks will be used. The output is compatible to be stored as alternative experiments.

altExps(se) <- splitByRanks(se)
se
## class: TreeSummarizedExperiment 
## dim: 19216 26 
## metadata(0):
## assays(1): counts
## rownames(19216): 549322 522457 ... 200359 271582
## rowData names(7): Kingdom Phylum ... Genus Species
## colnames(26): CL3 CC1 ... Even2 Even3
## colData names(7): X.SampleID Primer ... SampleType Description
## reducedDimNames(0):
## mainExpName: NULL
## altExpNames(7): Kingdom Phylum ... Genus Species
## rowLinks: a LinkDataFrame (19216 rows)
## rowTree: 1 phylo tree(s) (19216 leaves)
## colLinks: NULL
## colTree: NULL
altExpNames(se)
## [1] "Kingdom" "Phylum"  "Class"   "Order"   "Family"  "Genus"   "Species"

4.2 Constructing a tree from taxonomic data

Constructing a taxonomic tree from taxonomic data stored in rowData is quite straightforward and uses mostly functions implemented in TreeSummarizedExperiment.

taxa <- rowData(altExp(se,"Species"))[,taxonomyRanks(se)]
taxa_res <- resolveLoop(as.data.frame(taxa))
taxa_tree <- toTree(data = taxa_res)
taxa_tree$tip.label <- getTaxonomyLabels(altExp(se,"Species"))
rowNodeLab <- getTaxonomyLabels(altExp(se,"Species"), make_unique = FALSE)
altExp(se,"Species") <- changeTree(altExp(se,"Species"),
                                   rowTree = taxa_tree,
                                   rowNodeLab = rowNodeLab)

4.3 Transformation of assay data

Transformation of count data stored in assays is also a main task when work with microbiome data. transformCounts can be used for this and offers a few choices of available transformations. A modified object is returned and the transformed counts are stored in a new assay.

assayNames(enterotype)
## [1] "counts"
anterotype <- transformCounts(enterotype, method = "log10", pseudocount = 1)
assayNames(enterotype)
## [1] "counts"

For more details have a look at the man page ?transformCounts.

4.4 Community indices.

In the field of microbiome ecology several indices to describe samples and community of samples are available. In this vignette we just want to give a very brief introduction.

Functions for calculating alpha and beta diversity indices are available. Using estimateDiversity multiple diversity indices are calculated by default and results are stored automatically in colData. Selected indices can be calculated individually by setting index = "shannon" for example.

se <- estimateDiversity(se)
colnames(colData(se))[8:ncol(colData(se))]
## [1] "coverage"            "fisher"              "gini_simpson"       
## [4] "inverse_simpson"     "log_modulo_skewness" "shannon"            
## [7] "faith"

Beta diversity indices are used to describe inter-sample connections. Technically they are calculated as dist object and reduced dimensions can be extracted using cmdscale. This is wrapped up in the runMDS function of the scater package, but can be easily used to calculated beta diversity indices using the established functions from the vegan package or any other package using comparable inputs.

library(scater)
altExp(se,"Genus") <- runMDS(altExp(se,"Genus"), 
                             FUN = vegan::vegdist,
                             method = "bray",
                             name = "BrayCurtis", 
                             ncomponents = 5, 
                             exprs_values = "counts", 
                             keep_dist = TRUE)

JSD and UniFrac are implemented in mia as well. calculateJSD can be used as a drop-in replacement in the example above (omit the method argument as well) to calculate the JSD. For calculating the UniFrac distance via calculateUniFrac either a TreeSummarizedExperiment must be used or a tree supplied via the tree argument. For more details see ?calculateJSD, ?calculateUniFrac or ?calculateDPCoA.

runMDS performs the decomposition. Alternatively runNMDS can also be used.

4.5 Other indices

estimateDominance and estimateEvenness implement other sample-wise indices. The function behave equivalently to estimateDiversity. For more information see the corresponding man pages.

5 Utility functions

To make migration and adoption as easy as possible several utility functions are available.

5.1 Data loading functions

Functions to load data from biom files, QIIME2 output, DADA2 objects (Callahan et al. 2016) or phyloseq objects are available.

data(esophagus, package = "phyloseq")
esophagus
## phyloseq-class experiment-level object
## otu_table()   OTU Table:         [ 58 taxa and 3 samples ]
## phy_tree()    Phylogenetic Tree: [ 58 tips and 57 internal nodes ]
esophagus <- makeTreeSummarizedExperimentFromphyloseq(esophagus)
esophagus
## class: TreeSummarizedExperiment 
## dim: 58 3 
## metadata(0):
## assays(1): counts
## rownames(58): 59_8_22 59_5_13 ... 65_9_9 59_2_6
## rowData names(0):
## colnames(3): B C D
## colData names(0):
## reducedDimNames(0):
## mainExpName: NULL
## altExpNames(0):
## rowLinks: a LinkDataFrame (58 rows)
## rowTree: 1 phylo tree(s) (58 leaves)
## colLinks: NULL
## colTree: NULL

For more details have a look at the man page, for examples ?makeTreeSummarizedExperimentFromphyloseq.

5.2 General wrapper functions

getAbundanceFeature and getAbundanceSample are wrappers on row-wise or column-wise assay data subsetting.

abund <- getAbundanceSample(se, "CC1", abund_values = "counts")
all(abund == assay(se, "counts")[,"CC1"])
## [1] TRUE
abund <- getAbundanceFeature(se, "522457", abund_values = "counts")
all(abund == assay(se, "counts")["522457",])
## [1] TRUE

5.3 Selecting most interesting features

getTopTaxa returns a vector of the most top abundant feature IDs.

top_taxa <- getTopTaxa(se,
                       method = "mean",
                       top = 5,
                       abund_values = "counts")
top_taxa
## [1] "549656" "331820" "279599" "360229" "317182"

5.4 Generating tidy data

To generate tidy data as used and required in most of the tidyverse, meltAssay can be used. A data.frame in the long format will be returned.

molten_data <- meltAssay(se,
                         add_row_data = TRUE,
                         add_col_data = TRUE,
                         abund_values = "counts")
molten_data
## # A tibble: 499,616 × 24
##    FeatureID SampleID counts Kingdom Phylum   Class   Order Family Genus Species
##    <fct>     <fct>     <dbl> <chr>   <chr>    <chr>   <chr> <chr>  <chr> <chr>  
##  1 549322    CL3           0 Archaea Crenarc… Thermo… <NA>  <NA>   <NA>  <NA>   
##  2 549322    CC1           0 Archaea Crenarc… Thermo… <NA>  <NA>   <NA>  <NA>   
##  3 549322    SV1           0 Archaea Crenarc… Thermo… <NA>  <NA>   <NA>  <NA>   
##  4 549322    M31Fcsw       0 Archaea Crenarc… Thermo… <NA>  <NA>   <NA>  <NA>   
##  5 549322    M11Fcsw       0 Archaea Crenarc… Thermo… <NA>  <NA>   <NA>  <NA>   
##  6 549322    M31Plmr       0 Archaea Crenarc… Thermo… <NA>  <NA>   <NA>  <NA>   
##  7 549322    M11Plmr       0 Archaea Crenarc… Thermo… <NA>  <NA>   <NA>  <NA>   
##  8 549322    F21Plmr       0 Archaea Crenarc… Thermo… <NA>  <NA>   <NA>  <NA>   
##  9 549322    M31Tong       0 Archaea Crenarc… Thermo… <NA>  <NA>   <NA>  <NA>   
## 10 549322    M11Tong       0 Archaea Crenarc… Thermo… <NA>  <NA>   <NA>  <NA>   
## # … with 499,606 more rows, and 14 more variables: X.SampleID <fct>,
## #   Primer <fct>, Final_Barcode <fct>, Barcode_truncated_plus_T <fct>,
## #   Barcode_full_length <fct>, SampleType <fct>, Description <fct>,
## #   coverage <int>, fisher <dbl>, gini_simpson <dbl>, inverse_simpson <dbl>,
## #   log_modulo_skewness <dbl>, shannon <dbl>, faith <dbl>

6 Session info

sessionInfo()
## R version 4.1.0 (2021-05-18)
## Platform: x86_64-pc-linux-gnu (64-bit)
## Running under: Ubuntu 20.04.2 LTS
## 
## Matrix products: default
## BLAS:   /home/biocbuild/bbs-3.13-bioc/R/lib/libRblas.so
## LAPACK: /home/biocbuild/bbs-3.13-bioc/R/lib/libRlapack.so
## 
## locale:
##  [1] LC_CTYPE=en_US.UTF-8       LC_NUMERIC=C              
##  [3] LC_TIME=en_GB              LC_COLLATE=C              
##  [5] LC_MONETARY=en_US.UTF-8    LC_MESSAGES=en_US.UTF-8   
##  [7] LC_PAPER=en_US.UTF-8       LC_NAME=C                 
##  [9] LC_ADDRESS=C               LC_TELEPHONE=C            
## [11] LC_MEASUREMENT=en_US.UTF-8 LC_IDENTIFICATION=C       
## 
## attached base packages:
## [1] parallel  stats4    stats     graphics  grDevices utils     datasets 
## [8] methods   base     
## 
## other attached packages:
##  [1] phyloseq_1.36.0                scater_1.20.1                 
##  [3] ggplot2_3.3.5                  scuttle_1.2.0                 
##  [5] mia_1.0.8                      TreeSummarizedExperiment_2.0.2
##  [7] Biostrings_2.60.1              XVector_0.32.0                
##  [9] SingleCellExperiment_1.14.1    SummarizedExperiment_1.22.0   
## [11] Biobase_2.52.0                 GenomicRanges_1.44.0          
## [13] GenomeInfoDb_1.28.1            IRanges_2.26.0                
## [15] S4Vectors_0.30.0               BiocGenerics_0.38.0           
## [17] MatrixGenerics_1.4.0           matrixStats_0.60.0            
## [19] BiocStyle_2.20.2              
## 
## loaded via a namespace (and not attached):
##   [1] ggbeeswarm_0.6.0            colorspace_2.0-2           
##   [3] ellipsis_0.3.2              BiocNeighbors_1.10.0       
##   [5] bit64_4.0.5                 fansi_0.5.0                
##   [7] decontam_1.12.0             codetools_0.2-18           
##   [9] splines_4.1.0               sparseMatrixStats_1.4.0    
##  [11] cachem_1.0.5                knitr_1.33                 
##  [13] ade4_1.7-17                 jsonlite_1.7.2             
##  [15] cluster_2.1.2               BiocManager_1.30.16        
##  [17] compiler_4.1.0              assertthat_0.2.1           
##  [19] Matrix_1.3-4                fastmap_1.1.0              
##  [21] lazyeval_0.2.2              cli_3.0.1                  
##  [23] BiocSingular_1.8.1          htmltools_0.5.1.1          
##  [25] tools_4.1.0                 igraph_1.2.6               
##  [27] rsvd_1.0.5                  gtable_0.3.0               
##  [29] glue_1.4.2                  GenomeInfoDbData_1.2.6     
##  [31] reshape2_1.4.4              dplyr_1.0.7                
##  [33] Rcpp_1.0.7                  jquerylib_0.1.4            
##  [35] rhdf5filters_1.4.0          vctrs_0.3.8                
##  [37] multtest_2.48.0             ape_5.5                    
##  [39] nlme_3.1-152                DECIPHER_2.20.0            
##  [41] iterators_1.0.13            DelayedMatrixStats_1.14.0  
##  [43] xfun_0.24                   stringr_1.4.0              
##  [45] beachmat_2.8.0              lifecycle_1.0.0            
##  [47] irlba_2.3.3                 zlibbioc_1.38.0            
##  [49] MASS_7.3-54                 scales_1.1.1               
##  [51] biomformat_1.20.0           rhdf5_2.36.0               
##  [53] yaml_2.2.1                  memoise_2.0.0              
##  [55] gridExtra_2.3               sass_0.4.0                 
##  [57] stringi_1.7.3               RSQLite_2.2.7              
##  [59] foreach_1.5.1               ScaledMatrix_1.0.0         
##  [61] tidytree_0.3.4              permute_0.9-5              
##  [63] BiocParallel_1.26.1         rlang_0.4.11               
##  [65] pkgconfig_2.0.3             bitops_1.0-7               
##  [67] evaluate_0.14               lattice_0.20-44            
##  [69] Rhdf5lib_1.14.2             purrr_0.3.4                
##  [71] treeio_1.16.1               bit_4.0.4                  
##  [73] tidyselect_1.1.1            plyr_1.8.6                 
##  [75] magrittr_2.0.1              bookdown_0.22              
##  [77] R6_2.5.0                    generics_0.1.0             
##  [79] DelayedArray_0.18.0         DBI_1.1.1                  
##  [81] pillar_1.6.2                withr_2.4.2                
##  [83] mgcv_1.8-36                 survival_3.2-11            
##  [85] RCurl_1.98-1.3              tibble_3.1.3               
##  [87] crayon_1.4.1                utf8_1.2.2                 
##  [89] rmarkdown_2.9               viridis_0.6.1              
##  [91] grid_4.1.0                  data.table_1.14.0          
##  [93] blob_1.2.2                  vegan_2.5-7                
##  [95] digest_0.6.27               tidyr_1.1.3                
##  [97] munsell_0.5.0               DirichletMultinomial_1.34.0
##  [99] beeswarm_0.4.0              viridisLite_0.4.0          
## [101] vipor_0.4.5                 bslib_0.2.5.1

References

Amezquita, Robert, Aaron Lun, Etienne Becht, Vince Carey, Lindsay Carpp, Ludwig Geistlinger, Federico Marini, et al. 2020. “Orchestrating Single-Cell Analysis with Bioconductor.” Nature Methods 17: 137–45. https://www.nature.com/articles/s41592-019-0654-x.

Callahan, Benjamin J, Paul J McMurdie, Michael J Rosen, Andrew W Han, Amy Jo A Johnson, and Susan P Holmes. 2016. “DADA2: High-Resolution Sample Inference from Illumina Amplicon Data.” Nature Methods 13: 581–83. https://doi.org/10.1038/nmeth.3869.

Huang, Ruizhu. 2021. TreeSummarizedExperiment: TreeSummarizedExperiment: A S4 Class for Data with Tree Structures.

Morgan, Martin, Valerie Obenchain, Jim Hester, and Hervé Pagès. 2020. SummarizedExperiment: SummarizedExperiment Container. https://bioconductor.org/packages/SummarizedExperiment.