Last updated: 2022-07-09
Checks: 7 0
Knit directory: BASS-analysis/
This reproducible R Markdown analysis was created with workflowr (version 1.7.0). The Checks tab describes the reproducibility checks that were applied when the results were created. The Past versions tab lists the development history.
Great! Since the R Markdown file has been committed to the Git repository, you know the exact version of the code that produced these results.
Great job! The global environment was empty. Objects defined in the global environment can affect the analysis in your R Markdown file in unknown ways. For reproduciblity it’s best to always run the code in an empty environment.
The command set.seed(0) was run prior to running the
code in the R Markdown file. Setting a seed ensures that any results
that rely on randomness, e.g. subsampling or permutations, are
reproducible.
Great job! Recording the operating system, R version, and package versions is critical for reproducibility.
Nice! There were no cached chunks for this analysis, so you can be confident that you successfully produced the results during this run.
Great job! Using relative paths to the files within your workflowr project makes it easier to run your code on other machines.
Great! You are using Git for version control. Tracking code development and connecting the code version to the results is critical for reproducibility.
The results in this page were generated with repository version 8acc73b. See the Past versions tab to see a history of the changes made to the R Markdown and HTML files.
Note that you need to be careful to ensure that all relevant files for
the analysis have been committed to Git prior to generating the results
(you can use wflow_publish or
wflow_git_commit). workflowr only checks the R Markdown
file, but you know if there are other scripts or data files that it
depends on. Below is the status of the Git repository when the results
were generated:
working directory clean
Note that any generated files, e.g. HTML, png, CSS, etc., are not included in this status report because it is ok for generated content to have uncommitted changes.
These are the previous versions of the repository in which changes were
made to the R Markdown (analysis/about.Rmd) and HTML
(docs/about.html) files. If you’ve configured a remote Git
repository (see ?wflow_git_remote), click on the hyperlinks
in the table below to view the files as they were in that past version.
| File | Version | Author | Date | Message |
|---|---|---|---|---|
| Rmd | 8acc73b | zhengli09 | 2022-07-09 | Update introduction page |
| html | 303b575 | zhengli09 | 2022-06-12 | Build site. |
| Rmd | 829d10c | zhengli09 | 2022-06-12 | Update real data analysis |
| html | 7d4801a | zhengli09 | 2022-06-07 | Build site. |
| Rmd | 6a631f8 | zhengli09 | 2022-06-07 | Update software introduction and tutorial |
| html | 182beeb | zhengli09 | 2022-03-10 | Build site. |
| Rmd | 34e84d6 | zhengli09 | 2022-03-10 | Add software tutorial and multi-sample simulation |
| html | 9a68511 | zhengli09 | 2022-02-27 | Build site. |
| Rmd | 4ec8df1 | zhengli09 | 2022-02-27 | Publish the introduction, simulation analysis and STARmap analysis |
| html | b2c476b | zhengli09 | 2022-02-06 | Build site. |
| html | a6097f4 | zhengli09 | 2022-02-06 | Build site. |
| Rmd | 9f8756a | zhengli09 | 2022-02-06 | Start workflowr project. |
In this tutorial, we introduce the workflow and core functions of
BASS to help you get familiar with the BASS package and start analyzing
your own datasets. Please check the Real data analyses
sections for examples on how we analyze spatial transcriptomic
datasets.
BASS is implemented as an R (>= 4.0.3) package with underlying
efficient C++ code interfaced through Rcpp and RcppArmadillo. BASS
depends on a few other R packages that include GIGrvg, Matrix, harmony,
label.switching, mclust, Rcpp, RcppArmadillo, RcppDist, SPARK, scran,
and scater. Please refer to the package DESCRIPTION
file for details. Dependent packages are supposed to be automatically
installed while installing BASS. Install the BASS R package maintained
in github through the devtools package.
if(!require(devtools))
install.packages(devtools)
devtools::install_github("zhengli09/BASS")library(BASS)Find the data here.
load("data/starmap_mpfc.RData") # starmap_cnts, starmap_infoTo facilitate all steps of the spatial transcriptomic analysis, we
wrap up the raw data, processed data, model hyper-parameters, MCMC
controlling parameters, and model results into an S4 object
(BASS-class). Check details of the BASS-class
object by typing help("BASS-class"). Create a BASS object
to start your analysis with the function createBASSObject.
There are four required arguments: an L-list of gene expression count
matrices for L tissue sections (X), an L-list of spatial
coordinates matrices for L tissue sections (xy), the number
of cell types (C) and the number of spatial domains
(R). Check the simulation section of our paper for an
evaluation of the influence of the specified number of cell
types/spatial domains on the performance of all methods, and the
discussion section of our paper for a practical guideline on how to
select the two hyper-parameters when the truths are unclear. In
addition, when the purpose of the analysis is solely to detect spatial
domains, C can be specified to be a relatively large number
(e.g. C = 20) while exploring R. This ensures
that the heterogeneity in gene expression can be fully accounted. By
default, we infer the spatial interaction parameter \(\beta\) based on the data at hand by
specifying beta_method = "SW". However, if you have already
had a good sense about the value of \(\beta\) (usually between 1 and 2), you can
fix this parameter to the value specified by beta and set
beta_method = "fix". Check the documentation of
createBASSObject for further details.
Notes:
The BASS software currently implements a stopping rule to automatically
determine the number of MCMC samples required for inferring \(\beta\). In particular, we calculate the
mean of the sampled \(\beta\) in every
100 iterations and stop sampling if the difference of the consecutive
two means is below a certain threshold (tol = 0.001 by
default). Such stopping rule can help improve the computational
efficiency and works well in almost all cases. In cases where the
default number of MCMC iterations (2000) is not sufficient to reach the
stopping threshold, we have incorporated a warning message into the
software to encourage users to manually inspect the trace plot of \(\beta\) to determine if a larger number of
burn-in iterations is needed. Check the supplementary figure of our
paper for examples of well-behaved trace plots.
# gene expression count data
cnts <- starmap_cnts
# spatial coordinates
xy <- lapply(starmap_info, function(info.i){
as.matrix(info.i[, c("x", "y")])
})
BASS <- createBASSObject(cnts, xy, C = 15, R = 4, beta_method = "SW")***************************************
INPUT INFO:
- Number of tissue sections: 3
- Number of cells/spots: 1049 1053 1088
- Number of genes: 166
- Potts interaction parameter estimation method: SW
- Estimate Potts interaction parameter with SW algorithm
To list all hyper-parameters, Type listAllHyper(BASS_object)
***************************************You can check all the hyper-parameters that BASS uses with the
function listAllHyper.
listAllHyper(BASS)***************************************
Please refer to the paper for details of the paramters
ALL HYPER-PARAMETERS:
- Number of cell types C: 15
- Number of spatial domains R: 4
- Initialization method: kmeans
- Scale matrix of the inverse-Wishart prior Psi0: 1I
- Degrees of freedom of the inverse-Wishart prior n0: 1
- Number of MCMC burn-in iterations: 2000
- Number of MCMC posterior samples: 5000
- Potts interaction parameter estimation approach: SW
- Number of burn-in interations in Potts sampling: 10
- Number of Potts samples to approximate the partition ratio: 10
- Step size of a uniform random walk: 0.1
- Threshold of convergence for beta: 0.001
- Concentration parameter of the Dirichlet prior alpha0: 1
- Minimum number of neighbors for each cell/spot based on the Euclidean distance: 4
***************************************We pre-process the raw gene expression data stored at the
BASS@X slot and store the processed data at the
BASS@X_run slot for running the following BASS algorithm.
We have provided a function BASS.preprocess to automate the
pre-processing steps, where we combine cells/spots across all L tissue
sections, conduct library size normalization followed by a
log2-transformation (after adding a pseudo-count of 1), select feature
genes, perform dimension reduction with PCA on the normalized expression
matrix to extract J low-dimensional expression features, and perform
batch effect adjustment with Harmony to
align expression data from different tissue sections. Users can also
pre-process the gene expression data by themselves and store the
processed data to the BASS@X_run slot. Check the
documentation of BASS.preprocess for further details and
various other controls.
BASS <- BASS.preprocess(BASS)Run the Gibbs sampling algorithm in combination with the
Metropolis-Hastings algorithm to sample all the parameters specified in
the BASS model. Posterior samples of all the parameters are stored in
the BASS@samples slot and you can find a detailed
description of what has been stored in BASS@samples by
typing help("BASS-class"). Check the documentation of
BASS.run for further details.
BASS <- BASS.run(BASS)Post-process the posterior sampling results with the function
BASS.postprocess. There, we address the label switching
issue associated with the sampling of cell type labels and spatial
domain labels based on the iterative version 1 of the equivalence class
representation (ECR) algorithm implemented in the label.switching
package. In addition, we summarize posterior samples and obtain final
estimates of cell type labels, spatial domain labels, and cell type
composition in each spatial domain. The resulting estimates are stored
in the BASS@results slot. Check the documentation of
BASS.postprocess for further details.
Notes:
ECR-1 algorithm can be computationally intense when the sample size,
number of posterior samples, specified number of cell types, or
specified number of spatial domains is large. We provide the option,
adjustLS = FALSE, to directly obtain final estimates of the
cell type labels and spatial domain labels as the mode of their
posterior samples without adjusting for label switching. Such approach
can help improve the computational efficiency and the resulting
estimates are reasonably good.
BASS <- BASS.postprocess(BASS)
sessionInfo()R version 4.2.0 (2022-04-22)
Platform: x86_64-pc-linux-gnu (64-bit)
Running under: Ubuntu 18.04.5 LTS
Matrix products: default
BLAS: /usr/lib/x86_64-linux-gnu/openblas/libblas.so.3
LAPACK: /usr/lib/x86_64-linux-gnu/libopenblasp-r0.2.20.so
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=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] stats graphics grDevices utils datasets methods base
other attached packages:
[1] BASS_1.1.0 GIGrvg_0.5 workflowr_1.7.0
loaded via a namespace (and not attached):
[1] Rcpp_1.0.8.3 bslib_0.3.1 compiler_4.2.0 pillar_1.7.0
[5] later_1.1.0.1 git2r_0.28.0 jquerylib_0.1.4 tools_4.2.0
[9] getPass_0.2-2 mclust_5.4.9 digest_0.6.29 jsonlite_1.8.0
[13] evaluate_0.15 tibble_3.1.6 lifecycle_1.0.1 pkgconfig_2.0.3
[17] rlang_1.0.1 cli_3.2.0 rstudioapi_0.13 yaml_2.3.5
[21] xfun_0.29 fastmap_1.1.0 httr_1.4.2 stringr_1.4.0
[25] knitr_1.37 sass_0.4.1 fs_1.5.2 vctrs_0.3.8
[29] rprojroot_2.0.2 glue_1.6.2 R6_2.5.1 processx_3.5.2
[33] fansi_1.0.2 rmarkdown_2.12.1 callr_3.7.0 magrittr_2.0.2
[37] whisker_0.4 ps_1.6.0 promises_1.1.1 htmltools_0.5.2
[41] ellipsis_0.3.2 httpuv_1.5.4 utf8_1.2.2 stringi_1.7.6
[45] crayon_1.5.0