Microbiome data integration workflow for population cohort studies

Package demo

# List of packages that we need
packages <- c(
    "ANCOMBC", "ComplexHeatmap", "ggplot2", "knitr", "mia", "miaViz", "dplyr",
    "tidyr", "scater", "knitr")

# Get packages that are already installed installed
packages_already_installed <- packages[ packages %in% installed.packages() ]

# Get packages that need to be installed
packages_need_to_install <- setdiff( packages, packages_already_installed )

# Loads BiocManager into the session. Install it if it not already installed.
if( !require("BiocManager") ){
    install.packages("BiocManager")
    library("BiocManager")
}

# If there are packages that need to be installed, installs them with BiocManager
# Updates old packages.
if( length(packages_need_to_install) > 0 ) {
   install(packages_need_to_install, ask = FALSE)
}

# Load all packages into session. Stop if there are packages that were not
# successfully loaded
if( any(!sapply(packages, require, character.only = TRUE)) ){
    stop("Error in loading packages into the session.")
}

################################################################################
# Additional setup

# Set chunk options
opts_chunk$set(message = FALSE, warning = FALSE)

# Set black and white theme for figures, and Arial font
theme <- theme_bw() +
    theme(
        text = element_text(family = "Arial"), 
        panel.border = element_blank(), 
        panel.grid.major = element_blank(),
        panel.grid.minor = element_blank(), 
        axis.line = element_line(colour = "black")
        )
theme_set(theme)

EuroBioC2023

• European Bioconductor Conference 2023
• Ghent, Belgium
• September 21, 2023; 13:30 (CEST)
• See the poster also (miaverse – microbiome analytics framework in SummarizedExperiment family)!

Presenter information

All authors are affiliated to Turku Data Science Group in University of Turku, Finland.

• Tuomas Borman – doctoral researcher
• Chouaib Benchraka – doctoral researcher
• Leo Lahti – group leader

---

Learning goals

1. Microbiome research studies interactions between microbes (and human, environment…)
2. Big data requires efficient tools to manipulate the data
3. miaverse is a (Tree)SummarizedExperiment framework for microbiome analytics

Figure source: Moreno-Indias et al. (2021) Statistical and Machine Learning Techniques in Human Microbiome Studies: Contemporary Challenges and Solutions. Frontiers in Microbiology 12:11.

Motivation

Microbiome research

• Microbiome is a composition of microbes in well-defined area (gut, skin, mouth…)
• Bilateral interaction between human and microbiome –> affects both health and disease.
• The research is based on sequencing (characterization of genes and species).
• Nowadays, multiomics approach is more common (integration of taxonomy information with metabolite data, for example)
• Computational methods are the new microscope
• The research has expanded rapidly in previous years

# Plot publication graph
path <- "data/PubMed_Timeline_Results_by_Year.csv"
df <- read.csv(path, skip = 1)

x <- "Year"
y <- "Count"

plot <- ggplot(df, aes(x = .data[[x]], y = .data[[y]])) +
    geom_bar(stat="identity")
plot

PubMed publications per year with a search term ‘microbiome’ (fetched: Sep 5, 2023)

Big data

• Cohort datasets are large in size
  • Data management, handling and wrangling –> data structure
  • Computational power –> High performance computing (HPC) and cloud computing
• MultiAssayExperiment (MAE) and SummarizedExperiment (SE)
  • Several R packages frameworks are increasingly integrating MAE and SE
  • MAE enables linking of multiple experiments
  • SE – and especially TreeSE – is an efficient data container to store data from an experiment

miaverse (MIcrobiome Analysis)

• A framework for microbiome analytics
  • mia (analysis)
  • miaViz (visualization)
  • miaSim (simulation)
  • Orchestrating Microbiome Analysis, OMA (tutorial book)
• Based on TreeSummarizedExperiment (TreeSE) class
  • Supports hierarchical data
  • phyloseq class is a subset of TreeSE
  • Extension of SingleCellExperiment (SCE) class
  • miaverse is compatible with other SummarizedExperiment frameworks (especially scater package)
• Purpose to offer tools and tips for microbiome data analysis
• Enables development of versatile analytical workflows in microbiome data science
  • Supports multiomics (MultiAssayExperiment class)
  • Scalable
  • Standardized

The structure of the TreeSummarizedExperiment (TreeSE) class.

The workflow

The workflow is based on Orchestrating Microbiome Analysis (OMA) tutorial book. Find more information from there.

Importing the dataset

We fetch the data from MGnify database. It is a EMBL-EBI’s database for metagenomic data. This large microbiome database can be accessed with MGnifyR package which nowadays support TreeSE. The package will be submitted to Bioconductor’s next release.

We chose dataset of study MGYS00005128. In this study, they studied antibiotic resistance. They collected data from Cambodia, Kenya and UK. The dataset contains total of 1197 samples with taxonomy and gene function prediction data.

As loading takes some time, the dataset is already loaded.

# library(MGnifyR)
# # Create a client object
# mg <- MgnifyClient(useCache = TRUE, cacheDir = "data/magnifyr_cache")
# # Search analysis IDs based on study ID 
# analyses <- searchAnalysis(mg, "studies", "MGYS00005128")
# # Fetch data
# mae <- getResult(mg, analyses, get.func = "go-slim")
# # Store the data
# saveRDS(mae, "data/mae.Rds")

mae <- readRDS("data/mae.Rds")

The data containers

MAE stores multiple experiments, in this case 2 (taxonomy and gene function prediction info).

mae

## A MultiAssayExperiment object of 2 listed
##  experiments with user-defined names and respective classes.
##  Containing an ExperimentList class object of length 2:
##  [1] microbiota: TreeSummarizedExperiment with 2207 rows and 1197 columns
##  [2] go-slim: TreeSummarizedExperiment with 116 rows and 1197 columns
## Functionality:
##  experiments() - obtain the ExperimentList instance
##  colData() - the primary/phenotype DataFrame
##  sampleMap() - the sample coordination DataFrame
##  `$`, `[`, `[[` - extract colData columns, subset, or experiment
##  *Format() - convert into a long or wide DataFrame
##  assays() - convert ExperimentList to a SimpleList of matrices
##  exportClass() - save data to flat files

We can have general information on samples of the study in sample metadata of MAE.

colData(mae)[1:5, 1:5] %>% kable()

	analysis_accession	analysis_analysis.status	analysis_experiment.type	analysis_pipeline.version	analysis_is.private
MGYA00383606	MGYA00383606	completed	metagenomic	4.1	FALSE
MGYA00383607	MGYA00383607	completed	metagenomic	4.1	FALSE
MGYA00383608	MGYA00383608	completed	metagenomic	4.1	FALSE
MGYA00383609	MGYA00383609	completed	metagenomic	4.1	FALSE
MGYA00383610	MGYA00383610	completed	metagenomic	4.1	FALSE

MAE and TreeSE objects have rows and columns. This means that we can subset the data similarly to other objects that have rows and columns (like data.frame). In MAE, experiments and samples are linked together, meaning that we can subset the data at one go.

For demonstrative purpose and for saving resources, let’s subset the data by taking 100 random samples.

set.seed(49585)
random_samples <- sample(colnames(mae[[1]]), 100)
mae <- mae[, random_samples]
mae

## A MultiAssayExperiment object of 2 listed
##  experiments with user-defined names and respective classes.
##  Containing an ExperimentList class object of length 2:
##  [1] microbiota: TreeSummarizedExperiment with 2207 rows and 100 columns
##  [2] go-slim: TreeSummarizedExperiment with 116 rows and 100 columns
## Functionality:
##  experiments() - obtain the ExperimentList instance
##  colData() - the primary/phenotype DataFrame
##  sampleMap() - the sample coordination DataFrame
##  `$`, `[`, `[[` - extract colData columns, subset, or experiment
##  *Format() - convert into a long or wide DataFrame
##  assays() - convert ExperimentList to a SimpleList of matrices
##  exportClass() - save data to flat files

The first experiment / TreeSE includes taxonomy information.

mae[[1]]

## class: TreeSummarizedExperiment 
## dim: 2207 100 
## metadata(0):
## assays(1): counts
## rownames(2207): 148939 125998 ... 233398 233398.1
## rowData names(7): Kingdom Phylum ... Genus Species
## colnames(100): MGYA00393990 MGYA00384766 ... MGYA00384647 MGYA00394128
## colData names(64): analysis_accession analysis_analysis.status ...
##   sample_instrument.model sample_last.update.date
## reducedDimNames(0):
## mainExpName: NULL
## altExpNames(0):
## rowLinks: NULL
## rowTree: NULL
## colLinks: NULL
## colTree: NULL

The second one includes gene function prediction data. As you can see, we can fetch the data by specifying index or name of experiment.

mae[["go-slim"]]

## class: TreeSummarizedExperiment 
## dim: 116 100 
## metadata(0):
## assays(1): counts
## rownames(116): GO:0009317 GO:0016597 ... GO:0019012 GO:0019842
## rowData names(3): description category index_id
## colnames(100): MGYA00393990 MGYA00384766 ... MGYA00384647 MGYA00394128
## colData names(64): analysis_accession analysis_analysis.status ...
##   sample_instrument.model sample_last.update.date
## reducedDimNames(0):
## mainExpName: NULL
## altExpNames(0):
## rowLinks: NULL
## rowTree: NULL
## colLinks: NULL
## colTree: NULL

Taxonomy information includes phylogenetic table in its feature metadata.

rowData(mae[[1]]) %>% head() %>% kable()

	Kingdom	Phylum	Class	Order	Family	Genus	Species
148939	Eukaryota	Ascomycota	Dothideomycetes	NA	NA	NA	NA
125998	Eukaryota	Ascomycota	Eurotiomycetes	Eurotiales	NA	NA	NA
114164	Eukaryota	Ascomycota	Eurotiomycetes	Onygenales	NA	NA	NA
76021	Eukaryota	Ascomycota	Eurotiomycetes	NA	NA	NA	NA
73314	Eukaryota	Ascomycota	Saccharomycetes	Saccharomycetales	Debaryomycetaceae	Candida	NA
73314.1	Eukaryota	Ascomycota	Saccharomycetes	Saccharomycetales	Debaryomycetaceae	Candida	NA

Compared to phyloseq object, TreeSE can hold more data, for example, multiple assays. Let’s transform the data. Transformed table is stored to assays slot.

mae[[1]] <- transformAssay(mae[[1]], method = "relabundance")
mae[[1]]

## class: TreeSummarizedExperiment 
## dim: 2207 100 
## metadata(0):
## assays(2): counts relabundance
## rownames(2207): 148939 125998 ... 233398 233398.1
## rowData names(7): Kingdom Phylum ... Genus Species
## colnames(100): MGYA00393990 MGYA00384766 ... MGYA00384647 MGYA00394128
## colData names(64): analysis_accession analysis_analysis.status ...
##   sample_instrument.model sample_last.update.date
## reducedDimNames(0):
## mainExpName: NULL
## altExpNames(0):
## rowLinks: NULL
## rowTree: NULL
## colLinks: NULL
## colTree: NULL

Summarize data

We can summarize how many unique bacteria there in in certain taxonomy levels. For instance, we can see that there are 53 unique bacterial phyla.

rowData(mae[[1]]) %>% as_tibble() %>% summarise_all(n_distinct) %>% kable()

Kingdom	Phylum	Class	Order	Family	Genus	Species
5	53	108	175	273	685	882

A common operation in microbiome data analysis is agglomeration. This means that we sum-up the data to certain taxonomy levels. We can use mia::mergeFeaturesByRank function for agglomerating data to single taxonomy level. If we want to agglomerate the data to all found taxonomy levels with one command, we can use mia::splitByRanks.

altExp slot is the right place to store experiments with modified features (such as agglomerated or subsetted data).

altExps(mae[[1]]) <- splitByRanks(mae[[1]])
mae[[1]]

## class: TreeSummarizedExperiment 
## dim: 2207 100 
## metadata(0):
## assays(2): counts relabundance
## rownames(2207): 148939 125998 ... 233398 233398.1
## rowData names(7): Kingdom Phylum ... Genus Species
## colnames(100): MGYA00393990 MGYA00384766 ... MGYA00384647 MGYA00394128
## colData names(64): analysis_accession analysis_analysis.status ...
##   sample_instrument.model sample_last.update.date
## reducedDimNames(0):
## mainExpName: NULL
## altExpNames(7): Kingdom Phylum ... Genus Species
## rowLinks: NULL
## rowTree: NULL
## colLinks: NULL
## colTree: NULL

We can fetch agglomerated data from the slot. We can see that instead of 2207 features, there is only 5 features in the data that is summed-up to kingdom level.

altExp(mae[[1]], "Kingdom")

## class: TreeSummarizedExperiment 
## dim: 5 100 
## metadata(1): agglomerated_by_rank
## assays(2): counts relabundance
## rownames(5): Eukaryota Bacteria Archaea Chloroplast Mitochondria
## rowData names(7): Kingdom Phylum ... Genus Species
## colnames(100): MGYA00393990 MGYA00384766 ... MGYA00384647 MGYA00394128
## colData names(64): analysis_accession analysis_analysis.status ...
##   sample_instrument.model sample_last.update.date
## reducedDimNames(0):
## mainExpName: NULL
## altExpNames(0):
## rowLinks: NULL
## rowTree: NULL
## colLinks: NULL
## colTree: NULL

assays include agglomerated abundance tables.

assay(altExp(mae[[1]], "Kingdom"), "counts") %>% head() %>% kable()

	MGYA00393990	MGYA00384766	MGYA00384868	MGYA00393846	MGYA00393894	MGYA00393985	MGYA00384631	MGYA00394152	MGYA00383695	MGYA00384928	MGYA00383726	MGYA00393933	MGYA00385075	MGYA00384842	MGYA00384761	MGYA00384939	MGYA00393750	MGYA00384640	MGYA00384849	MGYA00385081	MGYA00394120	MGYA00384927	MGYA00384959	MGYA00384916	MGYA00384698	MGYA00384790	MGYA00393938	MGYA00385072	MGYA00394132	MGYA00394013	MGYA00384733	MGYA00393782	MGYA00383709	MGYA00385144	MGYA00393804	MGYA00383747	MGYA00385128	MGYA00394130	MGYA00393978	MGYA00385042	MGYA00384881	MGYA00393796	MGYA00384645	MGYA00384786	MGYA00384812	MGYA00384996	MGYA00383614	MGYA00393840	MGYA00383719	MGYA00393998	MGYA00393725	MGYA00394160	MGYA00384995	MGYA00383665	MGYA00383617	MGYA00384754	MGYA00394066	MGYA00393893	MGYA00393808	MGYA00394134	MGYA00385032	MGYA00386685	MGYA00393923	MGYA00384685	MGYA00384848	MGYA00385096	MGYA00393743	MGYA00393916	MGYA00384817	MGYA00393793	MGYA00394171	MGYA00384965	MGYA00384705	MGYA00393732	MGYA00384735	MGYA00384712	MGYA00384869	MGYA00383622	MGYA00384669	MGYA00384852	MGYA00384966	MGYA00394027	MGYA00384709	MGYA00393848	MGYA00385039	MGYA00385048	MGYA00383727	MGYA00394061	MGYA00393948	MGYA00384784	MGYA00393958	MGYA00383773	MGYA00384974	MGYA00393889	MGYA00384771	MGYA00383733	MGYA00384764	MGYA00394040	MGYA00384647	MGYA00394128
Eukaryota	0	11	78	17	1097	0	3	0	0	103	197	20	1515	38	5	2	18	71	9	1540	14	52	2	10062	37	9	28	1926	105	1316	3	912	2	925	1025	0	1098	252	151	1865	0	162	1874	3269	0	1812	4	14	2554	14	1164	23	1768	1776	2	0	1422	1046	1108	7	2841	157	36	1086	1	9	603	32	11	2131	2281	38	1	10	776	7	49	0	1676	5	3	0	0	3	17	0	178	33	53	3173	20	75	35	1057	0	83	0	249	2647	12
Bacteria	12972	24905	17949	22423	2542	13278	15987	11276	11418	15904	15870	13654	2925	31623	27460	26498	12378	16092	20844	2838	20757	9656	14791	95482	47183	25661	1011	4836	24770	15616	69616	21336	13801	13600	229	20324	9394	15200	16097	470	27951	12324	444	171780	24618	12197	27097	22691	1601	11288	3088	17242	12101	8357	27102	27228	457	2622	736	47864	2633	28480	29674	782	21101	16914	2601	50402	20847	6258	1805	11263	29342	17726	36623	43031	21714	28696	399	20574	14501	16719	28936	19623	32607	26047	16203	26306	58927	174135	38109	56421	10730	2759	22200	17384	25083	15233	2140	14658
Archaea	0	0	7	0	0	0	0	0	0	10	0	0	0	0	0	0	0	13	0	0	0	10	0	4	0	0	0	0	0	0	0	0	19	0	0	26	0	0	0	0	0	0	0	27	0	0	0	0	0	2	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	30	0	0	8	0	0	0	0	0	0	0	61	0	0	10	0	51	0	11	0	0	0	0	0	0	0	0
Chloroplast	76	264	102	289	3	95	0	95	37	4	105	236	7	0	14	268	182	3	71	9	57	79	165	189	34	318	14	2	101	143	490	66	63	4	0	208	25	128	118	1	151	8	1	1239	269	17	11	272	0	156	16	0	27	22	8	320	2	1	9	20	3	0	99	8	171	113	8	419	18	4	1	165	238	116	30	0	237	327	1	141	173	47	324	307	209	0	98	90	457	1119	13	177	134	3	197	215	32	124	6	13
Mitochondria	0	0	0	0	0	0	0	0	0	0	0	0	2	0	0	0	0	0	0	0	0	0	0	6	1	0	0	0	0	0	0	1	0	0	0	0	0	0	0	3	0	2	0	2	0	2	0	0	0	0	0	0	2	0	0	0	0	0	0	0	0	3	0	0	0	0	0	1	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	2	0	0	0	2	0	0	1	0	0	0	0	0	0	0	0

To visualize phylogenetic relations, we can first create a phylogenetic tree based on rowData, and then plot it. The tree is created in family level.

altExp(mae[[1]], "Family") <- addTaxonomyTree(altExp(mae[[1]], "Family"))
plotRowTree(altExp(mae[[1]], "Family"), edge_colour_by = "Kingdom")

miaverse includes several visualizing methods. For example, we can visualize relative abundances of 10 most abundant phyla.

plotAbundanceDensity(
    altExp(mae[[1]], "Phylum"), assay.type = "relabundance", n = 10, layout="density") 

Alpha diversity

Alpha diversity measures how diverse the microbial composition is. For example, if there are lots of different bacterial species, the alpha diversity is higher. Again, miaverse includes convenient tools to calculate alpha diversities of samples and to visualize them.

Here we analyze if alpha diversities differ between locations.

# Calculate
mae[[1]] <- estimateDiversity(mae[[1]], index = "shannon")
# Plot
plotColData(mae[[1]], y = "shannon", x = "location", colour_by = "location") +
    # These are normal ggplot objects
    theme(legend.position = "none")

Beta diversity

Beta diversity measures differences of microbial profiles of samples. There are several techniques Principal Component Analysis (PCA) being the most well-known ordination method. Distance-based redundancy analysis (dbRDA) is supervised ordination method which takes into account sample metadata. It maximizes the variance with respect to covariates.

Here we analyze if location can explain the differences in microbial profile. The result is stored to reducedDim slot.

mae[[1]] <- runRDA(
    mae[[1]],
    assay.type = "relabundance",
    formula = data ~ location,
    distance = "bray",
    name = "dbRDA"
)

reducedDim(mae[[1]], "dbRDA")

##                     dbRDA1       dbRDA2
## MGYA00393990  0.1347102942 -0.100832593
## MGYA00384766  0.1440370073 -0.133240761
## MGYA00384868  0.0752795935 -0.183802481
## MGYA00393846  0.0610872412  0.489951835
## MGYA00393894 -0.2329823176  0.069469681
## MGYA00393985  0.1341586408 -0.102957754
## MGYA00384631 -0.0316655436  0.277311863
## MGYA00394152  0.0618287158  0.493236009
## MGYA00383695 -0.0084027652 -0.155762139
## MGYA00384928 -0.1166574929 -0.068720426
## MGYA00383726  0.0736662587 -0.174236886
## MGYA00393933  0.0595845903  0.484813425
## MGYA00385075 -0.2476202356  0.089064859
## MGYA00384842  0.0560987204  0.084441075
## MGYA00384761  0.1335693958 -0.111399651
## MGYA00384939  0.1233061411 -0.107712350
## MGYA00393750  0.0613077131  0.491512394
## MGYA00384640 -0.1152849683 -0.055104759
## MGYA00384849  0.0691988641 -0.118067879
## MGYA00385081 -0.2435778106  0.081108922
## MGYA00394120  0.0727140079 -0.115199175
## MGYA00384927  0.0744038815 -0.184540075
## MGYA00384959  0.1227612375 -0.105682000
## MGYA00384916 -0.1450605401 -0.082901182
## MGYA00384698  0.0621041269 -0.046376995
## MGYA00384790  0.1440894238 -0.132580603
## MGYA00393938  0.0773613625 -0.117812352
## MGYA00385072 -0.2463605951  0.075185591
## MGYA00394132  0.0125192818 -0.092162835
## MGYA00394013  0.0949048574 -0.121352348
## MGYA00384733  0.1299577328 -0.107163273
## MGYA00393782 -0.0485745141 -0.084911171
## MGYA00383709  0.0344791254 -0.176607956
## MGYA00385144 -0.1743700351  0.025745252
## MGYA00393804 -0.0331379977  0.245074548
## MGYA00383747  0.1202622083 -0.160957854
## MGYA00385128 -0.1157261802  0.095426271
## MGYA00394130  0.1100743500 -0.151624819
## MGYA00393978  0.0749827716 -0.169128101
## MGYA00385042 -0.2386491739  0.176320933
## MGYA00384881  0.0703782568 -0.165098363
## MGYA00393796 -0.0931293914 -0.092467612
## MGYA00384645 -0.2301181982  0.164162146
## MGYA00384786  0.0502844922 -0.178037602
## MGYA00384812  0.1399074536 -0.126607636
## MGYA00384996 -0.1351452341 -0.109831900
## MGYA00383614 -0.0345784398  0.244892145
## MGYA00393840  0.0613128092  0.491673873
## MGYA00383719 -0.2079946631  0.087238702
## MGYA00393998  0.1397657271 -0.138923923
## MGYA00393725 -0.0696509924 -0.075937955
## MGYA00394160  0.0549978102  0.086559521
## MGYA00384995 -0.1318146008 -0.110735038
## MGYA00383665 -0.1308114996 -0.066265707
## MGYA00383617 -0.0342930884  0.243270312
## MGYA00384754  0.1444104462 -0.125846612
## MGYA00394066 -0.1934873125  0.101619718
## MGYA00393893 -0.2295337408  0.068759432
## MGYA00393808 -0.1755577610  0.005031351
## MGYA00394134  0.1397203992 -0.099679138
## MGYA00385032 -0.2486498398  0.124041724
## MGYA00386685 -0.1269555631  0.021395117
## MGYA00393923  0.1109984687 -0.138860763
## MGYA00384685 -0.1657696868 -0.005621145
## MGYA00384848  0.1051087071 -0.180194961
## MGYA00385096  0.0766841236 -0.038968087
## MGYA00393743 -0.1505195024 -0.077945676
## MGYA00393916  0.1487237066 -0.107600058
## MGYA00384817  0.0534617372 -0.053099266
## MGYA00393793 -0.2454584161  0.014680854
## MGYA00394171 -0.2101295794  0.092617510
## MGYA00384965  0.1375433669 -0.142346565
## MGYA00384705  0.1408101483 -0.097320538
## MGYA00393732  0.0184656582 -0.121993265
## MGYA00384735  0.0596693137  0.487015833
## MGYA00384712  0.0117835216  0.054146791
## MGYA00384869  0.1280221883 -0.153727071
## MGYA00383622  0.1403129292 -0.128121361
## MGYA00384669 -0.2029823451  0.082198544
## MGYA00384852  0.1046645459 -0.179625391
## MGYA00384966  0.1238179369 -0.108690300
## MGYA00394027  0.1098098043 -0.067429451
## MGYA00384709  0.1403300420 -0.127865221
## MGYA00393848  0.0617271155  0.491612219
## MGYA00385039  0.0195541731 -0.123350724
## MGYA00385048 -0.0313709212  0.271745946
## MGYA00383727  0.0722204721 -0.173850173
## MGYA00394061  0.0431684905 -0.156035621
## MGYA00393948  0.1136389418 -0.053411435
## MGYA00384784  0.0502472886 -0.180846880
## MGYA00393958 -0.0257144179  0.088235614
## MGYA00383773 -0.0563405106 -0.121503598
## MGYA00384974  0.1354616969 -0.143851679
## MGYA00393889 -0.2240017574  0.064431657
## MGYA00384771  0.0618245875  0.493131802
## MGYA00383733  0.1343592838 -0.124630454
## MGYA00384764  0.0606454204  0.487786298
## MGYA00394040  0.1113961726 -0.153128594
## MGYA00384647 -0.2406707164  0.050088550
## MGYA00394128 -0.0009264312 -0.084710131
## attr(,"rotation")
##      [,1] [,2]
## attr(,"eigen")
##   dbRDA1   dbRDA2 
## 4.116854 1.767239 
## attr(,"rda")
## Call: vegan::dbrda(formula = data ~ location, data = variables,
## distance = "bray")
## 
##               Inertia Proportion Rank RealDims
## Total         28.2023     1.0000              
## Constrained    5.8841     0.2086    2        2
## Unconstrained 22.3182     0.7914   97       61
## Inertia is squared Bray distance 
## Species scores projected from 'species_scores' 
## 
## Eigenvalues for constrained axes:
## dbRDA1 dbRDA2 
##  4.117  1.767 
## 
## Eigenvalues for unconstrained axes:
##  MDS1  MDS2  MDS3  MDS4  MDS5  MDS6  MDS7  MDS8 
## 6.778 3.098 2.828 1.530 1.322 0.991 0.787 0.694 
## (Showing 8 of 97 unconstrained eigenvalues)
## 
## attr(,"significance")
## attr(,"significance")$permanova
##          Df  SumOfSqs       F Pr(>F) Total variance Explained variance
## Model     2  5.884093 12.7868  0.001       28.20231          0.2086387
## location  2  5.884093 12.7868  0.001       28.20231          0.2086387
## Residual 97 22.318218      NA     NA       28.20231          0.7913613
## 
## attr(,"significance")$homogeneity
##          Df    Sum Sq   Mean Sq        F N.Perm Pr(>F) Total variance
## location  2 0.7177062 0.3588531 10.34989    999  0.001       4.080905
##          Explained variance
## location          0.1758694

As we can see, location has significant effect on microbial profile. However, from above we can see that groups do not have similar variance which is an assumption of PERMANOVA. This has to be taken into account when making conclusions.

plotRDA(mae[[1]], "dbRDA", colour_by = "location")

Differential abundance analysis (DAA)

The idea of DAA is to analyze, if there are bacteria whose abundance differ between groups. There are multiple methods to test this (such as basic Wilcoxon test). ANCOM-BC is a method that takes into consideration unique characters and features of microbial data.

Here we want to test if there are phyla whose abundance differ between locations.

# Analyze
res <- ancombc2(
    data =  altExp(mae[[1]], "Phylum"),
    fix_formula = "location",
    p_adj_method = "fdr",
    group = "location",
    global = TRUE
)
# Store results to data container
metadata( altExp(mae[[1]], "Phylum") )[["ancombc2"]] <- res
# Print
temp <- res$res_global
temp %>% kable()

taxon	W	p_val	q_val	diff_abn	passed_ss
Ascomycota	3.0152035	0.1303474	0.2929859	FALSE	TRUE
Basidiomycota	2.7706441	0.1637526	0.2932449	FALSE	FALSE
Arthropoda	1.7377258	0.3792163	0.4634866	FALSE	FALSE
Chordata	29.2649292	0.0000000	0.0000000	TRUE	TRUE
Nematoda	2.4336285	0.2096921	0.3075484	FALSE	FALSE
Acidobacteria	38.9192976	0.0000382	0.0002805	TRUE	TRUE
Actinobacteria	0.7400422	0.9602141	0.9602141	FALSE	FALSE
Bacillariophyta	1.4432923	0.5973256	0.6570582	FALSE	FALSE
Bacteroidetes	14.0562019	0.0000090	0.0000988	TRUE	TRUE
Chloroflexi	2.1208994	0.3413546	0.4417530	FALSE	FALSE
Cyanobacteria	19.4555055	0.0016916	0.0062024	TRUE	FALSE
Deinococcus-Thermus	2.4445394	0.1866104	0.2932449	FALSE	FALSE
Euryarchaeota	4.5780354	0.0625114	0.1719062	FALSE	FALSE
Fibrobacteres	1.6473489	0.4169444	0.4827777	FALSE	FALSE
Firmicutes	2.7863339	0.1331754	0.2929859	FALSE	FALSE
Fusobacteria	2.0794228	0.2643005	0.3634131	FALSE	TRUE
Gemmatimonadetes	6.3421884	0.0100882	0.0317058	TRUE	FALSE
Lentisphaerae	0.8346123	0.9195272	0.9602141	FALSE	FALSE
Proteobacteria	10.8483507	0.0001120	0.0006160	TRUE	TRUE
Spirochaetes	3.1806770	0.1803801	0.2932449	FALSE	TRUE
Tenericutes	2.8137936	0.1488332	0.2932449	FALSE	FALSE
Verrucomicrobia	9.6943325	0.0007929	0.0034888	TRUE	TRUE

# Add results to feature metadata
# Ensure that results go to right feature
rownames(temp) <- temp$taxon
temp <- temp[rownames(altExp(mae[[1]], "Phylum")), ]
rownames(temp) <- rownames( altExp(mae[[1]], "Phylum") )
# Add to rowData
rowData(altExp(mae[[1]], "Phylum")) <- cbind(rowData(altExp(mae[[1]], "Phylum")), temp)

We can visualize statistically significant features with boxplot.

# Get the data from assay, rowData and colData
df <- meltAssay(altExp(mae[[1]], "Phylum"), assay.type = "relabundance", add_col_data = TRUE, add_row_data = TRUE)
# Take only significant features
df <- df[ df$diff_abn, ]
# Plot
ggplot(df, aes(x = location, colour = location, y = relabundance)) +
    geom_boxplot(outlier.shape = NA) + 
    geom_jitter(width = 0.2) +
    # Own panel for each feature
    facet_grid(cols = vars(FeatureID)) +
    # Remove x axis text
    theme(axis.title.x=element_blank(), axis.text.x=element_blank()) +
    # Logarithmic scale
    scale_y_log10()

Cross-correlation

To demonstrate, how we can integrate experiments, we perform simple cross-correlation analysis. The purpose is to analyze, if there are phyla whose abundance correlates with predicted gene functions.

First we subset the gene function prediction data by taking only those features whose abundance varies the most across samples.

# Transform assay
mae[[2]] <- transformAssay(mae[[2]], method = "log10", pseudocount = 1)
mae[[2]] <- transformAssay(mae[[2]], assay.type = "log10", method = "z", name = "log10_z")
# Get coefficients of variances
rowData(mae[[2]])[["cv"]] <- apply( assay(mae[[2]], "log10_z"), 1, function(x) sd(x)/mean(x) )
# Subset the data by taking top 40 features
top_feat <- order(abs(rowData(mae[[2]])[["cv"]]), decreasing = TRUE)[1:40]
altExp(mae[[2]], "sub") <- mae[[2]][top_feat, ]
# Replace feature names with more desriptive names
rownames(altExp(mae[[2]], "sub") ) <- rowData(altExp(mae[[2]], "sub"))[["description"]]
# Print
altExp(mae[[2]], "sub")

## class: TreeSummarizedExperiment 
## dim: 40 100 
## metadata(0):
## assays(3): counts log10 log10_z
## rownames(40): cell adhesion transcription factor binding ...
##   sporulation iron-sulfur cluster binding
## rowData names(4): description category index_id cv
## colnames(100): MGYA00393990 MGYA00384766 ... MGYA00384647 MGYA00394128
## colData names(64): analysis_accession analysis_analysis.status ...
##   sample_instrument.model sample_last.update.date
## reducedDimNames(0):
## mainExpName: NULL
## altExpNames(0):
## rowLinks: NULL
## rowTree: NULL
## colLinks: NULL
## colTree: NULL

# Transform assay of microbial data
altExp(mae[[1]], "Phylum") <- transformAssay(altExp(mae[[1]], "Phylum"), method = "clr", pseudocount = 1)
# Perform cross-correlation analysis
res <- testExperimentCrossAssociation(
    mae,
    experiment1 = 1, experiment2 = 2,
    altexp1 = "Phylum", altexp2 = "sub",
    assay.type1 = "clr", assay.type = "log10_z",
    mode = "matrix"
)
# Store the result to data container
metadata(mae)[["croscor"]] <- res
# Plot
plot <- Heatmap(
    res$cor, name = "Kendall's tau",
    # Print values to cells
    cell_fun = function(j, i, x, y, width, height, fill) {
        # If the p-value is under threshold
        if( !is.na(res$p_adj[i, j]) & res$p_adj[i, j] < 0.05 ){
            # Print "X"
            grid.text(sprintf("%s", "X"), x, y, gp = gpar(fontsize = 8, col = "black"))
            }
        },
    column_names_rot = 45
    )
# Adjust padding around plot so that names are visible
draw(plot, padding = unit(c(10, 40, 2, 2), "mm"))

Save the results

Finally, we can save the data container which contains our analysis results.

saveRDS(mae, "data/mae_results.Rds")

Thank you for your time!

Key points

1. Microbiome research studies interactions between microbes (and human, environment…)
2. Big data requires efficient tools to manipulate the data
3. miaverse is a (Tree)SummarizedExperiment framework for microbiome analytics

More information

• Project website
• Poster: miaverse – microbiome analytics framework in SummarizedExperiment family

Session info

sessionInfo()

## R version 4.3.1 (2023-06-16)
## Platform: x86_64-pc-linux-gnu (64-bit)
## Running under: Linux Mint 21
## 
## Matrix products: default
## BLAS:   /opt/R/4.3.1/lib/R/lib/libRblas.so 
## LAPACK: /usr/lib/x86_64-linux-gnu/lapack/liblapack.so.3.10.0
## 
## locale:
##  [1] LC_CTYPE=en_US.UTF-8       LC_NUMERIC=C              
##  [3] LC_TIME=en_US.UTF-8        LC_COLLATE=en_US.UTF-8    
##  [5] LC_MONETARY=fi_FI.UTF-8    LC_MESSAGES=en_US.UTF-8   
##  [7] LC_PAPER=fi_FI.UTF-8       LC_NAME=C                 
##  [9] LC_ADDRESS=C               LC_TELEPHONE=C            
## [11] LC_MEASUREMENT=fi_FI.UTF-8 LC_IDENTIFICATION=C       
## 
## time zone: Europe/Helsinki
## tzcode source: system (glibc)
## 
## attached base packages:
## [1] stats4    grid      stats     graphics  grDevices utils     datasets 
## [8] methods   base     
## 
## other attached packages:
##  [1] doRNG_1.8.6                    rngtools_1.5.2                
##  [3] foreach_1.5.2                  scater_1.29.4                 
##  [5] scuttle_1.11.2                 tidyr_1.3.0                   
##  [7] dplyr_1.1.3                    miaViz_1.9.3                  
##  [9] ggraph_2.1.0.9000              mia_1.9.16                    
## [11] MultiAssayExperiment_1.27.5    TreeSummarizedExperiment_2.9.0
## [13] Biostrings_2.69.2              XVector_0.41.1                
## [15] SingleCellExperiment_1.23.0    SummarizedExperiment_1.31.1   
## [17] Biobase_2.61.0                 GenomicRanges_1.53.1          
## [19] GenomeInfoDb_1.37.4            IRanges_2.35.2                
## [21] S4Vectors_0.39.1               BiocGenerics_0.47.0           
## [23] MatrixGenerics_1.13.1          matrixStats_1.0.0             
## [25] knitr_1.43                     ggplot2_3.4.3                 
## [27] ComplexHeatmap_2.17.0          ANCOMBC_2.3.1                 
## [29] BiocManager_1.30.22           
## 
## loaded via a namespace (and not attached):
##   [1] splines_4.3.1               ggplotify_0.1.2            
##   [3] bitops_1.0-7                tibble_3.2.1               
##   [5] cellranger_1.1.0            polyclip_1.10-4            
##   [7] rpart_4.1.19                DirichletMultinomial_1.43.0
##   [9] lifecycle_1.0.3             Rdpack_2.5                 
##  [11] doParallel_1.0.17           lattice_0.21-8             
##  [13] MASS_7.3-60                 backports_1.4.1            
##  [15] magrittr_2.0.3              Hmisc_5.1-0                
##  [17] sass_0.4.7                  rmarkdown_2.24             
##  [19] jquerylib_0.1.4             yaml_2.3.7                 
##  [21] gld_2.6.6                   cowplot_1.1.1              
##  [23] RColorBrewer_1.1-3          DBI_1.1.3                  
##  [25] minqa_1.2.5                 multcomp_1.4-25            
##  [27] abind_1.4-5                 zlibbioc_1.47.0            
##  [29] expm_0.999-7                purrr_1.0.2                
##  [31] RCurl_1.98-1.12             TH.data_1.1-2              
##  [33] yulab.utils_0.0.9           nnet_7.3-19                
##  [35] tweenr_2.0.2                sandwich_3.0-2             
##  [37] circlize_0.4.15             GenomeInfoDbData_1.2.10    
##  [39] ggrepel_0.9.3               irlba_2.3.5.1              
##  [41] tidytree_0.4.5              vegan_2.6-4                
##  [43] permute_0.9-7               DelayedMatrixStats_1.23.4  
##  [45] codetools_0.2-19            DelayedArray_0.27.10       
##  [47] ggforce_0.4.1               energy_1.7-11              
##  [49] shape_1.4.6                 tidyselect_1.2.0           
##  [51] aplot_0.2.0                 farver_2.1.1               
##  [53] lme4_1.1-34                 gmp_0.7-2                  
##  [55] ScaledMatrix_1.9.1          viridis_0.6.4              
##  [57] base64enc_0.1-3             jsonlite_1.8.7             
##  [59] GetoptLong_1.0.5            BiocNeighbors_1.19.0       
##  [61] e1071_1.7-13                tidygraph_1.2.3            
##  [63] decontam_1.21.0             Formula_1.2-5              
##  [65] survival_3.5-7              iterators_1.0.14           
##  [67] ggnewscale_0.4.9            tools_4.3.1                
##  [69] treeio_1.25.4               DescTools_0.99.50          
##  [71] Rcpp_1.0.11                 glue_1.6.2                 
##  [73] BiocBaseUtils_1.3.2         gridExtra_2.3              
##  [75] SparseArray_1.1.11          xfun_0.40                  
##  [77] mgcv_1.9-0                  withr_2.5.0                
##  [79] numDeriv_2016.8-1.1         fastmap_1.1.1              
##  [81] boot_1.3-28.1               bluster_1.11.4             
##  [83] fansi_1.0.4                 digest_0.6.33              
##  [85] rsvd_1.0.5                  gridGraphics_0.5-1         
##  [87] R6_2.5.1                    colorspace_2.1-0           
##  [89] Cairo_1.6-1                 gtools_3.9.4               
##  [91] RSQLite_2.3.1               utf8_1.2.3                 
##  [93] generics_0.1.3              data.table_1.14.8          
##  [95] DECIPHER_2.29.0             class_7.3-22               
##  [97] graphlayouts_1.0.0          CVXR_1.0-11                
##  [99] httr_1.4.7                  htmlwidgets_1.6.2          
## [101] S4Arrays_1.1.6              pkgconfig_2.0.3            
## [103] gtable_0.3.4                Exact_3.2                  
## [105] Rmpfr_0.9-3                 blob_1.2.4                 
## [107] htmltools_0.5.6             clue_0.3-64                
## [109] scales_1.2.1                lmom_3.0                   
## [111] png_0.1-8                   ggfun_0.1.2                
## [113] rstudioapi_0.15.0           rjson_0.2.21               
## [115] reshape2_1.4.4              checkmate_2.2.0            
## [117] nlme_3.1-163                nloptr_2.0.3               
## [119] GlobalOptions_0.1.2         zoo_1.8-12                 
## [121] proxy_0.4-27                cachem_1.0.8               
## [123] stringr_1.5.0               rootSolve_1.8.2.3          
## [125] parallel_4.3.1              vipor_0.4.5                
## [127] foreign_0.8-84              pillar_1.9.0               
## [129] vctrs_0.6.3                 BiocSingular_1.17.1        
## [131] beachmat_2.17.15            cluster_2.1.4              
## [133] beeswarm_0.4.0              htmlTable_2.4.1            
## [135] evaluate_0.21               magick_2.7.5               
## [137] mvtnorm_1.2-3               cli_3.6.1                  
## [139] compiler_4.3.1              rlang_1.1.1                
## [141] crayon_1.5.2                labeling_0.4.3             
## [143] plyr_1.8.8                  ggbeeswarm_0.7.2           
## [145] stringi_1.7.12              viridisLite_0.4.2          
## [147] BiocParallel_1.35.4         lmerTest_3.1-3             
## [149] munsell_0.5.0               gsl_2.1-8                  
## [151] lazyeval_0.2.2              Matrix_1.6-1               
## [153] patchwork_1.1.3             sparseMatrixStats_1.13.4   
## [155] bit64_4.0.5                 highr_0.10                 
## [157] rbibutils_2.2.15            igraph_1.5.1               
## [159] memoise_2.0.1               bslib_0.5.1                
## [161] ggtree_3.9.1                bit_4.0.5                  
## [163] readxl_1.4.3                ape_5.7-1