scRNA-Seq binarization with scBoolSeq

scRNA-Seq binarization with scBoolSeq#

The tool scBoolSeq can be used to binarize scRNA-Seq datasets according to gene-wise pseudocount distributions learnt from a reference scRNA-Seq dataset.

from scboolseq import scBoolSeq
import pandas as pd

import warnings
warnings.simplefilter(action='ignore', category=FutureWarning)

1. Learning of reference pseudocount distributions

scBoolSeq learns pseudocount distributions from highly-variables genes of a reference scRNA-Seq dataset.

!test -f GSE81682_Hematopoiesis.csv || curl -fOL \
    https://github.com/bnediction/scBoolSeq-supplementary/raw/main/data_filtered_vargenes/GSE81682_Hematopoiesis.csv

ref_data = pd.read_csv("GSE81682_Hematopoiesis.csv", index_col=0)
# for the sake of the demo, we shrink the number of genes
ref_data = ref_data.iloc[:,42:242]
ref_data

	2900041M22Rik	Klk8	Gm37637	Gp9	Idh3a	Akr1c13	2810408A11Rik	Npr2	Ephx1	Pik3ip1	...	Ctsl	Iigp1	P2ry14	Cd82	Slc18a2	Cd302	Parp12	Isyna1	S100a8	B130034C11Rik
HSPC_025	0.000000	0.000000	0.000000	1.832751	8.774285	1.189716	1.189716	1.832751	0.0	3.118770	...	1.832751	1.189716	7.790156	7.421084	0.000000	0.000000	6.779938	2.275971	6.079295	0.000000
HSPC_031	0.686872	0.686872	0.000000	0.000000	2.827390	8.409573	0.686872	1.782055	0.0	8.765434	...	0.000000	0.000000	8.900430	2.397399	1.500480	0.000000	0.686872	3.582099	0.000000	0.000000
HSPC_037	0.000000	7.275944	0.000000	0.000000	8.862937	0.000000	0.000000	1.218731	0.0	8.954567	...	7.160213	1.869807	10.783395	8.057904	0.000000	0.000000	6.235258	3.164226	0.000000	0.000000
LT-HSC_001	0.000000	7.520353	0.000000	0.000000	2.364517	0.000000	0.000000	3.217169	0.0	7.232592	...	8.058464	4.909226	3.749470	3.749470	0.000000	2.364517	8.700371	0.000000	0.000000	2.364517
HSPC_001	0.000000	4.414804	0.377367	7.107768	8.010188	5.867946	0.000000	9.303379	0.0	8.824449	...	8.121239	0.676211	8.514683	1.762030	0.377367	0.000000	0.000000	1.628905	0.377367	0.000000
...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...
Prog_834	5.354924	5.205256	5.661187	6.153235	7.324615	5.824095	0.000000	0.488810	0.0	5.464396	...	7.791800	6.454615	6.605636	6.581538	7.992470	2.642176	5.020429	6.042844	0.000000	4.965622
Prog_840	0.000000	5.009981	5.770268	0.470364	7.716333	5.780421	0.000000	0.470364	0.0	5.790503	...	6.429388	6.605327	6.486320	6.152863	6.390145	1.345870	0.470364	7.106828	0.000000	0.000000
Prog_846	1.024409	1.024409	0.601281	0.601281	7.614214	5.350986	0.000000	0.000000	0.0	4.133070	...	6.809524	0.601281	4.215689	3.042448	2.499161	0.000000	0.601281	7.156591	4.293831	0.000000
Prog_852	0.000000	0.000000	0.000000	0.000000	7.665835	6.581736	1.624428	0.000000	0.0	2.857900	...	2.369158	0.000000	4.448371	5.227124	1.029699	0.000000	2.044323	7.334552	4.963769	0.000000
Prog_810	0.329207	7.215698	0.597074	0.000000	7.464188	6.855898	0.000000	0.000000	0.0	7.404325	...	1.190055	0.000000	3.556499	4.748924	1.190055	0.000000	0.329207	1.725482	2.838296	4.374125

1656 rows × 200 columns

%time scbool = scBoolSeq().fit(ref_data)

Computing bimodality index for 89/200 genes
Computing bimodality index for 139/200 genes
CPU times: user 5.99 s, sys: 582 ms, total: 6.57 s
Wall time: 1.89 s

2. (optional) Access to the learnt pseudocount distributions

scbool.criteria_[['Category', *scbool.criteria_]]

	Category	Mean	MeanNZ	Median	MedianNZ	GeometricMean	HarmonicMean	Variance	VarianceNZ	DropOutRate	Amplitude	Dip	Kurtosis	Skewness	DenPeak	BI	Category
2900041M22Rik	ZeroInf	0.336663	2.844458	0.000000	1.434885	1.799128	1.053994	1.538907	5.868876	0.881643	9.051052	0.586866	17.547098	4.240360	0.004969	0.000000	ZeroInf
Klk8	Bimodal	3.284505	4.368787	1.742776	5.107305	2.969196	1.640595	10.048549	8.628783	0.248188	9.538356	0.000000	-1.530713	0.390004	0.373447	2.919056	Bimodal
Gm37637	ZeroInf	0.260708	2.585227	0.000000	1.232339	1.499722	0.910336	1.248158	6.367534	0.899155	8.323065	0.955873	26.252765	5.111231	0.004177	0.000000	ZeroInf
Gp9	Bimodal	1.674165	3.376878	0.000000	1.967766	2.150308	1.288318	6.875951	8.119299	0.504227	10.033402	0.000000	0.878241	1.530903	0.030183	2.745009	Bimodal
Idh3a	Bimodal	6.243458	6.370404	7.497524	7.541152	5.551172	4.379675	7.536478	6.881015	0.019928	9.882836	0.000000	-0.897883	-0.709034	8.245655	2.228428	Bimodal
...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...
Cd302	ZeroInf	0.527272	2.359899	0.000000	1.281987	1.474975	0.904786	2.130893	5.212369	0.776570	8.945604	0.199354	11.506878	3.443353	0.015439	0.000000	ZeroInf
Parp12	Bimodal	1.831223	3.434322	0.405306	1.998044	2.149494	1.244920	7.447516	8.461700	0.466787	10.432446	0.000900	0.585548	1.424771	0.056362	2.519183	Bimodal
Isyna1	Bimodal	5.861969	6.014511	7.093528	7.152238	5.082111	3.756957	8.006213	7.297088	0.025362	10.269083	0.000000	-1.108027	-0.601258	7.928887	2.419874	Bimodal
S100a8	ZeroInf	0.710715	3.155347	0.000000	1.983719	2.039669	1.233042	3.424959	7.492051	0.774758	16.763306	0.735221	11.336264	3.220410	0.021852	0.000000	ZeroInf
B130034C11Rik	ZeroInf	0.544910	2.570856	0.000000	1.591329	1.661072	1.029217	2.138895	4.882779	0.788043	8.442426	0.696348	9.344361	3.137202	0.011575	0.000000	ZeroInf

200 rows × 17 columns

3. Binarize data

Here, we binarize the reference dataset:

%time bindata = scbool.binarize(ref_data)
bindata

CPU times: user 94.6 ms, sys: 1.05 ms, total: 95.6 ms
Wall time: 94.9 ms

	2900041M22Rik	Klk8	Gm37637	Gp9	Idh3a	Akr1c13	2810408A11Rik	Npr2	Ephx1	Pik3ip1	...	Ctsl	Iigp1	P2ry14	Cd82	Slc18a2	Cd302	Parp12	Isyna1	S100a8	B130034C11Rik
HSPC_025	NaN	0.0	NaN	1.0	1.0	0.0	1.0	1.0	NaN	0.0	...	0.0	1.0	1.0	1.0	0.0	NaN	1.0	0.0	1.0	NaN
HSPC_031	1.0	0.0	NaN	0.0	0.0	1.0	1.0	1.0	NaN	1.0	...	0.0	NaN	1.0	0.0	0.0	NaN	1.0	0.0	NaN	NaN
HSPC_037	NaN	1.0	NaN	0.0	1.0	0.0	NaN	1.0	NaN	1.0	...	1.0	1.0	1.0	1.0	0.0	NaN	1.0	0.0	NaN	NaN
LT-HSC_001	NaN	1.0	NaN	0.0	0.0	0.0	NaN	1.0	NaN	1.0	...	1.0	1.0	NaN	0.0	0.0	1.0	1.0	0.0	NaN	1.0
HSPC_001	NaN	1.0	1.0	1.0	1.0	1.0	NaN	1.0	NaN	1.0	...	1.0	1.0	1.0	0.0	0.0	NaN	0.0	0.0	1.0	NaN
...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...
Prog_834	1.0	1.0	1.0	1.0	1.0	1.0	NaN	1.0	NaN	NaN	...	1.0	1.0	1.0	1.0	1.0	1.0	1.0	1.0	NaN	1.0
Prog_840	NaN	1.0	1.0	1.0	1.0	1.0	NaN	1.0	NaN	1.0	...	1.0	1.0	1.0	1.0	1.0	1.0	1.0	1.0	NaN	NaN
Prog_846	1.0	0.0	1.0	1.0	1.0	1.0	NaN	NaN	NaN	NaN	...	1.0	1.0	NaN	0.0	0.0	NaN	1.0	1.0	1.0	NaN
Prog_852	NaN	0.0	NaN	0.0	1.0	1.0	1.0	NaN	NaN	0.0	...	0.0	NaN	NaN	NaN	0.0	NaN	1.0	1.0	1.0	NaN
Prog_810	1.0	1.0	1.0	0.0	1.0	1.0	NaN	NaN	NaN	1.0	...	0.0	NaN	NaN	NaN	0.0	NaN	1.0	0.0	1.0	1.0

1656 rows × 200 columns