The orloj package includes a set of utilities for importing data from and interacting with the Astrolabe platform. As an Astrolabe user, you are welcome to download this data through the experiment Export button, which will download the experiment.zip file. For further description of the export file and the specifics of the Astrolabe pipeline, refer to the introduction.
You can install orloj through the R command line. Please make sure to run R in administrator or sudo mode, then execute the following:
- Install the required Bioconductor packages:
source("https://bioconductor.org/biocLite.R")
biocLite("flowCore")
biocLite("FlowSOM")
- Install devtools:
install.packages("devtools")
- Install orloj from the github repository:
devtools::install_github("astrolabediagnostics/orloj")
orloj provides access to the RDS files that are exported by Astrolabe. Before proceeding, please read the introduction to the various analyses and reports generated by Astrolabe.
Throughout this guide technical notes appear in blockquote.
The examples here use data from Tordesillas et al., J Allergy Clin Immunol, 2016. You can download the raw data from the FlowRepository repo, or download the Astrolabe
experiment.zipfile. We assume that you unzipped the file toD:/data/tordesillas.
TODO update link to file above
All orloj operations are done within the scope of an Astrolabe experiment. Start by loading the experiment:
> experiment <- orloj::loadExperiment([experiment_path])
To load the data used in this experiment, substitute [experiment_path] to the directory where you unzipped experiment.zip:
> experiment <- orloj::loadExperiment("D:/data/tordesillas")
The experiment directory will have the
config.RDSfile. This file includes the sample and feature information that was entered during Astrolabe experiment setup.
Use orloj::experimentSummary to get a brief summary of the experiment contents:
> orloj::experimentSummary(experiment)
Astrolabe experiment with 54 samples and 31 channels
Classification channels: CD66b, CD19, CD45RA, CD141, CD4, CD8a, CD20, CD16, CD66a, CD61, CD123, CD23, CD27, CD11c, CD14, CD32, CD63, CD3, CD38, CD56, HLADR
Sample features:
patient: c1, c2, c3, p1, p2, p3, p4, p5, p6
type: healthy, peanut allergic
condition: anti ige, media, peanut extract
timepoint: 15m, 30m
experiment is an R list which contains the parameters you supplied when setting up the experiment, in addition to various parameters created by Astrolabe. For example, you can access the experiment samples list using experiment$samples:
> print(experiment$samples)
# A tibble: 54 x 3
SampleId Name Filename
<chr> <chr> <chr>
1 1 C1 15min antiIgE C1_15min_antiIgE.fcs
2 2 C2 15min media C2_15min_media.fcs
3 3 C1 30min antiIgE C1_30min_antiIgE.fcs
4 4 C1 30min PN C1_30min_PN.fcs
5 5 C1 30min media C1_30min_media.fcs
6 6 C1 15min PN C1_15min_PN.fcs
7 7 C2 30min antiIgE C2_30min_antiIgE.fcs
8 8 C2 15min antiIgE C2_antiIgE_15min.fcs
9 9 C2 15min PN C2_15min_PN.fcs
10 10 C2 30min PN C2_30min_PN.fcs
# ... with 44 more rows
The SampleId is an internal Astrolabe field -- you will notice that the /analysis/ directory has several files of the format [sample_id].[analysis_name].RDS. Fortunately, you do not need to interact with these files directly. For a given sample, you can load all of the analyses for that sample:
> sample <- orloj::loadSample(experiment, sample_name = "C1 15min antiIgE")
> orloj::sampleSummary(sample)
An Astrolabe sample with 115242 cells and 49 channels
2661 bead events and 346946 debris events
If you would like to load samples using Astrolabe sample IDs, use
orloj::loadSampleData(experiment, sample_id = 1).
After loading a sample, you can access its FCS data and cell subset assignments:
> exprs <- orloj::fcsExprs(sample)
> dim(exprs)
[1] 115242 54
> colnames(exprs)
[1] "Time" "Event_length" "Pd102Di" "Pd104Di" "Pd105Di" "Pd106Di"
[7] "Pd108Di" "Pd110Di" "CD66b" "HLA_ABC" "Xe131Di" "Cs133Di"
[13] "Ce140Di" "Pr141Di" "CD19" "Ce142Di" "CD45RA" "CD141"
[19] "CD4" "CD8a" "CD20" "CD16" "CD66a" "CD61"
[25] "CD123" "Sm152Di" "CD23" "CD45" "CD27" "p38"
[31] "CD11c" "CD14" "CD32" "CRTH2" "IkBa" "FceRIa"
[37] "CD63" "Er167Di" "Tm169Di" "CD3" "pERK" "CD38"
[43] "CD56" "HLADR" "pS6" "Lu176Di" "Os189Di" "Ir191Di_DNA"
[49] "Ir193Di_DNA" "Assignment" "Level_0" "Level_1" "Level_2" "Profile"
exprs is a data frame with a row for each cell. It includes the original FCS data after transformation. Additionally, it has additional columns for Astrolabe's automatic cell subset classifier (gating). Assignment and Profile correspond to the terminal labels For each cell. The Level_i columns correspond to intermediate steps in the hierarchy -- we recommend against using them, unless you would like to focus on a specific subset.
Astrolabe uses hyperbolic arcsine with a cofactor of 5 when transforming mass cytometry data.
By default,
orloj::fcsExprsremoves events that were identified as debris by Astrolabe. If you would like to keep them, callorloj::fcsExprs(s, keep_debrus = TRUE)instead.orloj::fcsExprsalways removes events that were identified as CyTOF calibration beads.
orloj supplies two functions for easily accessing the sample's subet cell counts and channel intensity medians:
> sampleCellSubsetCounts(sample)
# A tibble: 12 x 2
CellSubset N
<chr> <int>
1 Basophil 2219
2 B Cell 211
3 CD141+ Conventional Dendritic Cell 714
4 CD14+ Monocyte 20945
5 CD16+ Monocyte 652
6 CD16- NK Cell 612
7 CD4+ T Cell 37815
8 CD8+ T Cell 21437
9 Granulocyte 16924
10 Myeloid_unassigned 4502
11 Root_unassigned 3297
12 T Cell_unassigned 5914
>
> sampleCellSubsetChannelStatistics(sample) %>% dplyr::arrange(CellSubset)
# A tibble: 336 x 5
CellSubset ChannelName Mean Sd Median
<chr> <chr> <dbl> <dbl> <dbl>
1 Basophil CD66b 1.0648623 0.8579865 0.9551815
2 Basophil HLA_ABC 3.5587049 0.6760540 3.5720787
3 Basophil CD19 0.3356665 0.5086242 0.0000000
4 Basophil CD45RA 0.6741676 0.6827153 0.5181072
5 Basophil CD141 0.3759006 0.5336876 0.0181990
6 Basophil CD4 0.1399595 0.3510070 0.0000000
7 Basophil CD8a 0.2580162 0.4292534 0.0000000
8 Basophil CD20 0.5901105 0.5963551 0.4241660
9 Basophil CD16 1.3112481 1.1552829 1.0417074
10 Basophil CD66a 2.5811629 1.6747920 2.2227098
# ... with 326 more rows
The subset cell counts over all samples can be accessed using orloj::experimentCellSubsetCounts:
> experimentCellSubsetCounts(experiment) %>% dplyr::arrange(CellSubset)
# A tibble: 689 x 5
SampleId Name Filename CellSubset N
<chr> <chr> <chr> <chr> <int>
1 1 C1 15min antiIgE C1_15min_antiIgE.fcs Basophil 2219
2 2 C2 15min media C2_15min_media.fcs Basophil 2916
3 3 C1 30min antiIgE C1_30min_antiIgE.fcs Basophil 2321
4 4 C1 30min PN C1_30min_PN.fcs Basophil 1943
5 5 C1 30min media C1_30min_media.fcs Basophil 2565
6 6 C1 15min PN C1_15min_PN.fcs Basophil 2438
7 7 C2 30min antiIgE C2_30min_antiIgE.fcs Basophil 3382
8 8 C2 15min antiIgE C2_antiIgE_15min.fcs Basophil 4949
9 9 C2 15min PN C2_15min_PN.fcs Basophil 3036
10 10 C2 30min PN C2_30min_PN.fcs Basophil 2881
# ... with 679 more rows
And similarly for subset channel intensity statistics:
> experimentCellSubsetChannelStatistics(experiment) %>% dplyr::arrange(CellSubset)
# A tibble: 19,292 x 8
SampleId Name Filename CellSubset ChannelName Mean Sd Median
<chr> <chr> <chr> <chr> <chr> <dbl> <dbl> <dbl>
1 1 C1 15min antiIgE C1_15min_antiIgE.fcs Basophil CD66b 1.0648623 0.8579865 0.9551815
2 1 C1 15min antiIgE C1_15min_antiIgE.fcs Basophil HLA_ABC 3.5587049 0.6760540 3.5720787
3 1 C1 15min antiIgE C1_15min_antiIgE.fcs Basophil CD19 0.3356665 0.5086242 0.0000000
4 1 C1 15min antiIgE C1_15min_antiIgE.fcs Basophil CD45RA 0.6741676 0.6827153 0.5181072
5 1 C1 15min antiIgE C1_15min_antiIgE.fcs Basophil CD141 0.3759006 0.5336876 0.0181990
6 1 C1 15min antiIgE C1_15min_antiIgE.fcs Basophil CD4 0.1399595 0.3510070 0.0000000
7 1 C1 15min antiIgE C1_15min_antiIgE.fcs Basophil CD8a 0.2580162 0.4292534 0.0000000
8 1 C1 15min antiIgE C1_15min_antiIgE.fcs Basophil CD20 0.5901105 0.5963551 0.4241660
9 1 C1 15min antiIgE C1_15min_antiIgE.fcs Basophil CD16 1.3112481 1.1552829 1.0417074
10 1 C1 15min antiIgE C1_15min_antiIgE.fcs Basophil CD66a 2.5811629 1.6747920 2.2227098
# ... with 19,282 more rows
Remember that you can always supply
level = "Profile"as a function parameter to access theProfilelevel instead of the defaultAssignmentlevel.
Finally, you can access the differential abundance analysis results:
> daa <- orloj::differentialAbundanceAnalysis(experiment)
> names(daa)
[1] "patient" "type" "condition" "timepoint"
Astrolabe runs the analysis for each of the features you supplied during experiment setup. Each of these include the P-values, FDR, and feature value medians for each cell subset:
> daa$condition
> daa$condition
CellSubset PValue FDR median_anti ige median_media median_peanut extract
1 T Cell_unassigned 0.1248214 0.9058288 0.0506369973 0.0493739530 4.847519e-02
2 Root_unassigned 0.2907044 0.9058288 0.0056848426 0.0056272949 5.843414e-03
3 Granulocyte 0.4423689 0.9058288 0.1092210834 0.0830771861 1.365757e-01
4 Plasmacytoid Dendritic Cell 0.4483833 0.9058288 0.0048429066 0.0042607659 4.857075e-03
5 CD4+ T Cell 0.5454668 0.9058288 0.2989413963 0.3207483271 3.111650e-01
6 CD8+ T Cell 0.5575078 0.9058288 0.1819280851 0.1982661806 1.906808e-01
7 B Cell 0.6167987 0.9058288 0.0448034615 0.0340524660 4.073054e-02
8 CD16+ Monocyte 0.7515697 0.9058288 0.0106419480 0.0108713938 1.169913e-02
9 CD14+ Monocyte 0.7838902 0.9058288 0.1333721631 0.1580286985 1.458530e-01
10 Basophil 0.7944606 0.9058288 0.0233674522 0.0246680809 2.144130e-02
11 CD16- NK Cell 0.7958336 0.9058288 0.0069009109 0.0055712110 6.395491e-03
12 Myeloid_unassigned 0.8383330 0.9058288 0.0498024335 0.0457445240 4.446624e-02
13 CD16+ NK Cell 0.8411267 0.9058288 0.0002580689 0.0003215008 8.402538e-05
14 CD141+ Conventional Dendritic Cell 0.9503421 0.9503421 0.0020229537 0.0018712104 2.828895e-03