scBoolSeq

scRNA-Seq data binarisation and synthetic generation from Boolean dynamics.

Installation

We recommend installing scBoolSeq via conda, but we provide as well a standard pip installation (which requires installing R and a set of R packages beforehand).

Conda

conda install -c conda-forge -c colomoto scboolseq

Pip

You need R installed, see the specification of the R dependencies below.

pip install scboolseq

Docker

The scBoolSeq command can be invoked using the bnediction/scboolseq image:

docker run --rm -it -v $PWD:/data -w /data bnediction/scboolseq scBoolSeq ...

Usage

Command line

scBoolSeq provides a rich CLI allowing programmatic access to its main functionalities, namely the binarization of RNA-Seq data and the generation of synthetic RNA-Seq data synthesis reflecting activation states from Boolean Network simulations. Once correctly instaled, the tool's and subcommand's help explain all the possible parameters. Some minimal examples are here presented.

Main CLI

$ scBoolSeq -h
usage: scBoolSeq <command> [<args>]

Available commands:
	* binarize	 Binarize a RNA-Seq dataset.
	* synthesize	 Simulate a RNA-Seq experiment from Boolean dynamics.
	* from_file	 Repeat a binarization or synthethic generation experiment, based on a config file.

NOTE on TSV/CSV file specs:
* If '.csv', the file is assumed to use the standard separator for columns ','.
* The index (gene or sample identifiers) is assumed to be the first column.
* The scBoolSeq is designed with consistency in mind. 
  The `output` (binarized or synthetic expression frame) will have the same disposition 
  (genes x observations | observations x genes) as the `input`. 
  If a `reference` is specified, its disposition must match the `input`'s.

scBoolSeq: bulk and single-cell RNA-Seq data binarization and synthetic generation from Boolean dynamics.

positional arguments:
  command     Subcommand to run

optional arguments:
  -h, --help  show this help message and exit

Binarization

Minimal example of binarization, specifying some optional parameters.

curl -fOL https://github.com/pinellolab/STREAM/raw/master/stream/tests/datasets/Nestorowa_2016/data_Nestorowa.tsv.gz

ls
# data_Nestorowa.tsv.gz
time scBoolSeq binarize data_Nestorowa.tsv.gz --genes-are-rows\
--output Nestorowa_binarized.csv --n-threads 10 --dump-config --dump-criteria
# ________________________________________________________
# Executed in   34.49 secs   fish           external 
#   usr time   30.04 secs  1211.00 micros   30.04 secs 
#   sys time    3.90 secs  171.00 micros    3.89 secs 

ls
# data_Nestorowa.tsv.gz    scBoolSeq_criteria_data_Nestorowa_2022-04-27_15h14m27.tsv
# Nestorowa_binarized.csv  scBoolSeq_experiment_config_2022-04-27_15h14m27.toml

# Visualize the binarized expression frame. 
# Note that some entries are undefined (NaN)
# These might be discarded genes for which no binarization or synthesis can occur,
# or observations which did not pass the thresholds to be set to 0 or 1.
python -c 'import pandas as pd; pd.read_csv("Nestorowa_binarized.csv", index_col=0).iloc[0:7, 0:7]'
#             Clec1b  Kdm3a  Coro2b  8430408G22Rik  Clec9a  Phf6  Usp14
# HSPC_025       NaN    1.0     NaN            NaN     NaN   0.0    0.0
# HSPC_031       NaN    1.0     NaN            NaN     NaN   0.0    0.0
# HSPC_037       NaN    0.0     1.0            NaN     NaN   0.0    1.0
# LT-HSC_001     NaN    0.0     1.0            NaN     NaN   1.0    0.0
# HSPC_001       NaN    0.0     1.0            NaN     NaN   1.0    0.0
# HSPC_008       1.0    1.0     NaN            NaN     NaN   1.0    0.0
# HSPC_014       NaN    0.0     NaN            NaN     NaN   0.0    1.0

Synthetic generation from Boolean states

cat minimal_boolean_example.csv 
# the output is not commented out so that it can be copied
# and perhaps be read with `x = pandas.read_clipboard(sep=',', index_col=0)`
,HSPC_025,HSPC_031,HSPC_037,LT-HSC_001,HSPC_001,HSPC_008,HSPC_014,HSPC_020,HSPC_026,HSPC_038,LT-HSC_002,HSPC_002,HSPC_009,HSPC_015,HSPC_021
Kdm3a,1.0,1.0,0.0,0.0,0.0,1.0,0.0,0.0,0.0,0.0,0.0,0.0,1.0,0.0,1.0
Coro2b,1.0,1.0,1.0,1.0,1.0,0.0,1.0,0.0,1.0,0.0,0.0,0.0,0.0,1.0,1.0
8430408G22Rik,1.0,0.0,0.0,1.0,0.0,1.0,1.0,1.0,1.0,0.0,0.0,0.0,1.0,0.0,1.0
Clec9a,1.0,0.0,0.0,1.0,1.0,0.0,1.0,0.0,1.0,1.0,0.0,0.0,0.0,0.0,0.0
Phf6,0.0,0.0,0.0,1.0,1.0,1.0,0.0,1.0,1.0,1.0,0.0,1.0,0.0,1.0,0.0


# Generate 20 samples per boolean state, using 12 threads
# setting the random number generator's seed ensures reproductiblility.
time scBoolSeq synthesize --genes-are-rows minimal_boolean_example_T.csv --reference data_Nestorowa.tsv.gz\
--n-samples 20 --output new_synthetic.tsv --n-threads 12 --rng-seed 1234
# ________________________________________________________
# Executed in   43.85 secs   fish           external 
#    usr time   22.08 secs    0.00 millis   22.08 secs 
#    sys time    3.65 secs    3.31 millis    3.65 secs 

# visualize the newly generated synthetic scRNA-Seq experiment
python -c 'import pandas as pd; pd.read_csv("new_synthetic.tsv", index_col=0, sep="\t").iloc[0:3, 0:7]'
#                HSPC_025  HSPC_031  HSPC_037  LT-HSC_001  HSPC_001  HSPC_008  HSPC_014
# Kdm3a          7.328819  8.536391  0.000000    0.000000  0.821561  7.030519  1.891949
# Coro2b         0.000000  0.000000  6.457878    5.479887  0.000000  0.000000  5.503554
# 8430408G22Rik  0.000000  0.005110  0.000000    0.000000  0.000000  6.428994  0.000000

Python API

Here a minimal example is presented, using the same dataset as the CLI usage guide. For further information, please check the documentation.

import pandas as pd
from scboolseq import scBoolSeq

# read in the normalized expression data
nestorowa = pd.read_csv("data_Nestorowa.tsv.gz", index_col=0, sep="\t")
nestorowa.iloc[1:5, 1:5] 
#                HSPC_031  HSPC_037  LT-HSC_001  HSPC_001
# Kdm3a          6.877725  0.000000    0.000000  0.000000
# Coro2b         0.000000  6.913384    8.178374  9.475577
# 8430408G22Rik  0.000000  0.000000    0.000000  0.000000
# Clec9a         0.000000  0.000000    0.000000  0.000000
#
# NOTE : here, genes are rows and observations are columns

# scBoolSeq expects genes to be columns, thus we transpose the DataFrame.
scbool_nest = scBoolSeq(data=nestorowa.T, r_seed=1234)

##
## Binarization
##

scbool_nest.fit() # compute binarization criteria

binarized = scbool_nestorowa.binarize(nestorowa.T)
binarized.iloc[1:5, 1:5] 
#             Kdm3a  Coro2b  8430408G22Rik  Phf6
# HSPC_031      1.0     NaN            NaN   0.0
# HSPC_037      0.0     1.0            NaN   0.0
# LT-HSC_001    0.0     1.0            NaN   1.0
# HSPC_001      0.0     1.0            NaN   1.0


##
## Synthetic RNA-Seq generation from Boolean states
##

scbool_nestorowa.simulation_fit() # compute simulation criteria

# we generate Boolean states by randomly (equiprobably) binarize undetermined
# values from the previous binarization.
from scboolseq.simulation import random_nan_binariser
fully_bin = binarized.iloc[1:5, 1:5].pipe(random_nan_binariser) 
fully_bin 
#             Kdm3a  Coro2b  8430408G22Rik  Phf6
# HSPC_031      1.0     0.0            1.0   0.0
# HSPC_037      0.0     1.0            1.0   0.0
# LT-HSC_001    0.0     1.0            0.0   1.0
# HSPC_001      0.0     1.0            1.0   1.0

# create a synthetic frame, with two samples per boolean state,
# fixing the rng's seed for reproducibility
scbool_nestorowa.simulate(fully_bin, n_threads=4, seed=1234, n_samples=2) 
#               Kdm3a    Coro2b  8430408G22Rik      Phf6
# HSPC_031    7.328819  0.000000       8.087928  0.923352
# HSPC_037    1.003712  6.843611       7.003577  0.000000
# LT-HSC_001  0.000000  0.000000       0.000000  5.174053
# HSPC_001    1.672793  0.000000       0.000000  4.481709
# HSPC_031    8.536391  1.060373       0.000000  3.267464
# HSPC_037    1.055816  5.479887       0.000000  3.836276
# LT-HSC_001  0.000000  0.000000       0.000000  8.131221
# HSPC_001    2.451340  0.000000       0.000000  9.969012

Contributors

• Gustavo Magaña López