The Self-Assembling-Manifold (SAM) algorithm.

• numpy
• scipy
• pandas
• scikit-learn
• umap-learn
• numba
• anndata

• Plotting

  • plotly==4.0.0
  • ipythonwidgets
  • jupyter
  • colorlover
  • ipyevents

• Clustering

  • louvain
  • leidenalg
  • hdbscan
  • cython

• scanpy

SAM has been most extensively tested using python3.6 but presumably should work on python>=3.6. Python can be installed using Anaconda.

Download Anacodna from here: https://www.anaconda.com/download/

Create and activate a new environment with python3.6 as follows:

conda create -n environment_name python=3.6
conda activate environment_name

Having activated the environment, SAM can be downloaded from the PyPI repository using pip or, for the development version, downloaded from the github directly.

PIP install:

pip install sam-algorithm

Development version install:

git clone https://github.com/atarashansky/self-assembling-manifold.git
cd self-assembling-manifold
python setup.py install

For plotting, install matplotlib:

pip install matplotlib

For interactive data exploration (in the SAMGUI.py module), jupyter, ipythonwidgets, colorlover, ipyevents, and plotly are required. Install them in the previously made environment like so:

conda install -c conda-forge -c plotly jupyter ipywidgets plotly=4.0.0 colorlover ipyevents #plotly-orca psutil requests

If you use Jupyter Notebooks, these steps are not needed. If you would like to be able to run SAMGUI in JupyterLab, please do the following:

First install nodejs with: conda install nodejs

To enable ipythonwidgets in Jupyter lab, please run the following:

jupyter labextension install @jupyter-widgets/[email protected] --no-build
jupyter labextension install [email protected] --no-build
jupyter labextension install [email protected] --no-build
jupyter lab build

SAMGUI should now work in JupyterLab.

The SAM GUI interface can be run in Jupyer notebooks with the following:

from SAMGUI import SAMGUI
sam_gui = SAMGUI(sam) # sam is your SAM object
sam_gui.SamPlot

Please see the plotting tutorial for more information about the GUI interface.

Please see the Jupyter notebooks in the 'tutorial' folder for basic tutorials. If you installed a fresh environment, do not forget to install jupyter into that environment! Please run

pip install jupyter

in your conda environment. The tutorial assumes that all optional dependencies are installed.

There are a number of different ways to load data into the SAM object.

from SAM import SAM #import SAM
sam=SAM(counts=(matrix,geneIDs,cellIDs))
sam.preprocess_data() # log transforms and filters the data
sam.run() #run with default parameters
sam.scatter()

from SAM import SAM #import SAM
sam=SAM(counts=dataframe)
sam.preprocess_data() # log transforms and filters the data
sam.run() #run with default parameters
sam.scatter()

from SAM import SAM #import SAM
sam=SAM(counts=adata)
sam.preprocess_data() # log transforms and filters the data
sam.run() #run with default parameters
sam.scatter()

from SAM import SAM #import SAM
sam=SAM() #initialize SAM object
sam.load_data('/path/to/expression_data_file.csv') #load data from a csv file
#sam.load_data('/path/to/expression_data_file.txt', sep='\t') #load data from a txt file with tab delimiters
sam.preprocess_data() # log transforms and filters the data
sam.load_annotations('/path/to/annotations_file.csv')
sam.run()
sam.scatter()

If loading tabular data (e.g. from a csv), load_data by default saves the sparse data structure to a h5ad file in the same location as the tabular file for faster loading in subsequent analyses. This file can be loaded as:

from SAM import SAM #import SAM
sam=SAM() #initialize SAM object
sam.load_data('/path/to/h5ad_file.h5ad') #load data from a h5ad file
sam.preprocess_data() # log transforms and filters the data
sam.run()
sam.scatter()

If you wish to save the SAM outputs and filtered data, you can write sam.adata to a h5ad file as follows: sam.save_anndata(filename, data = 'adata').

If you would like to save the entire attirbute dictionary of a SAM object to a Pickle file: sam.save(filename.p)

To load these attributes:

sam = SAM()
sam.load(filename.p)

If using the SAM algorithm, please cite the following eLife paper: https://elifesciences.org/articles/48994

Tarashansky, A. J. et al. Self-assembling manifolds in single-cell RNA sequencing data. eLife 8, e48994 (2019).

As always, please submit a new issue if you would like to see any functionalities / convenience functions / etc added.