Skip to content

GLM_supervised

Jump to bottom
marekkokot edited this page Aug 4, 2025 · 2 revisions

How to run it

The Tu run GLM supervised test flag should be used as follows: --supervised_test_samplesheet. Here is the simple snippet with an example of how to run it for 10X data:\

git clone --recurse-submodules https://github.com/refresh-bio/SPLASH/
cd SPLASH
make -j
cd example
./download_10X.py
../bin/splash --technology 10x --supervised_test_samplesheet test_data_cell_barcode_samplesheet_GLM_supervised_test.tsv input-10X.txt

If compilation is not feasible, the above may be replaced with using a precompiled release (but the download link should be adjusted to the appropriate platform and version, check releases):

wget https://github.com/refresh-bio/SPLASH/releases/download/v2.11.6/splash-2.11.6.linux.x64.tar.gz
tar -xvf splash-2.11.6.linux.x64.tar.gz
cd example
./download_10X.py
../splash --technology 10x --supervised_test_samplesheet test_data_cell_barcode_samplesheet_GLM_supervised_test.tsv input-10X.txt

metadata file format:

non-10x data:

The only requirement for metadata file is that the first column and second columns must be for sample names and the assigned class/group/category/celltype to each sample. Also, the first row must be column names (column names can be anything). The two columns are subsequently renamed as "sample_name" and "group" in the script. Below is an example metadata file:

sample_name	type
SRR8788980	carcinoma
SRR8788981	melanoma
SRR8788982	lymphoma
SRR8788983	carcinoma
SRR8657060	carcinoma

10x data:

Meta data file for 10x data must have an extra column (second column) for cell barcodes. Other requirements are the same as non-10x data:

sample_name cell cell_type
S2	AAACCCAAGACACACG	AT2
S1	AAACCCAAGCCATGCC	NK
S1	AAACCCAAGGGAACAA	Mac
S4	AAACCCAAGGGTTGCA	NK

Output files:

The output directory for the GLM test is: directory/run_name_supervised_metadata/.

  • GLM_supervised_anchors.tsv: the output file containing anchors with non-zero GLM_coefficients whose largest GLM coefficient is greater than 1.
  • plots: this subfolder contains the plots for visualizing the anchors with metadata-dependent target fraction varaition. For each anchor, two plots will be generated in a pdf named as anchor sequence, a box plot for showing the fraction of target1 per sample grouped by the category and a scatterplot for showing the counts for target1 and target2 per sample colorcoded by the metadata category.

Contact

Please contact Roozbeh Dehghannasiri (rdehghan@stanford.edu).

Table of contents

Home

  1. Quick start
  2. Installation
    1. Precompiled binaries
    2. Docker container
    3. Singularity container
    4. Compile from sources
    5. Running the example
  3. Configuration
    1. Base configuration
    2. Filters and thresholds
    3. Additional output configuration
    4. Tuning statistics computation
    5. Technical and performance-related
    6. sc-SPLASH (10x or visium processing)
  4. Input and output
    1. Input format
    2. Understanding the output
    3. Additional output
  5. sc-SPLASH: barcoded scRNA-Seq and Spatial applications
    1. Installation
    2. Input format
    3. Quick start
    4. Parameters
    5. BKC
  6. Compactors
    1. Installation
    2. Toy examples
    3. Parameters
  7. Lookup table
    1. General idea
    2. Building the index
    3. Querying the index
    4. Extension
    5. Using transcriptomes files
    6. Supported input formats for a query
    7. Formatting the output
    8. Canonical k-mers
  8. GLM-based supervised test
  9. Applications and results interpretation
  10. Postprocessing
Clone this wiki locally