Prepare kinase-substrate dataset

from matplotlib import pyplot as plt
import pandas as pd
import numpy as np
import seaborn as sns
from scipy.stats import spearmanr, pearsonr

Large scale dataset

# large_scale data
LS= pd.read_excel('raw/large_scale_final2.xlsx')

LS['gene_site'] = LS['S_position']

LS = LS[['Kinase','gene_site','substrate']]

LS_info = pd.read_csv('raw/LS_info.csv')

LS = LS.merge(LS_info).drop(columns = ['count'])

PhosphoSitePlus dataset

KS = pd.read_csv('raw/Kinase_Substrate_Dataset_final.csv')

# only keep human kinase substrate pairs
KS = KS.query('KIN_ORGANISM=="human"')

# remove the substrate that does not have sty in the middle
KS = KS[KS.substrate.str[7].isin(['s','t','y'])].reset_index(drop=True)

KS['gene_site'] = KS.SUB_GENE.str.upper() + "_"+KS.SUB_MOD_RSD

KS = KS[['Kinase','gene_site','substrate','kinase_uniprot','kinase_paper']]

Combine

KS['source'] = 'pplus'

LS['source'] = 'large_scale'

df = pd.concat([KS,LS]).reset_index(drop=True)

df

	Kinase	gene_site	substrate	kinase_uniprot	kinase_paper	source
0	DYRK2	GLI2_S385	AEGLRPAsPLGLTQE	Q92630	DYRK2	pplus
1	DYRK2	SIAH2_S68	GGGAGPVsPQHHELT	Q92630	DYRK2	pplus
2	DYRK2	CARHSP1_S41	LRGNVVPsPLPtRRt	Q92630	DYRK2	pplus
3	DYRK2	DCX_S306	GPMRRSKsPADSANG	Q92630	DYRK2	pplus
4	DYRK2	CDC25A_S283	PErsQEEsPPGSTKr	Q92630	DYRK2	pplus
...	...	...	...	...	...	...
216464	SPHK1	HERC2_S2928	AAVPFLAsDNEEEED	Q9NYA1	NaN	large_scale
216465	SPHK1	PARG_Y832	DALHFRRyLDQFVPE	Q9NYA1	NaN	large_scale
216466	SPHK1	G6PI_T109	ALRNRsNtPILVDGK	Q9NYA1	NaN	large_scale
216467	PRPK	TITIN_S20761	DAERKSWsTVTTECs	Q96S44	PRPK	large_scale
216468	PRPK	TITIN_S20768	sTVTTECsKTSFRVA	Q96S44	PRPK	large_scale

216469 rows × 6 columns

# kinase without mapping
df.loc[df.kinase_paper.isna()].Kinase.value_counts()

Kinase
LYNb            1694
ABL1[T315I]     1320
ABL1[E255K]     1274
RET[M918T]      1221
FGFR3[K650M]    1220
                ... 
GRP78              1
TAO3               1
PIK3CD             1
PIK3CB             1
G11                1
Name: count, Length: 80, dtype: int64

df.substrate.str[7].value_counts()

substrate
y    109225
s     74830
t     32414
Name: count, dtype: int64

df.source.value_counts()

source
large_scale    198248
pplus           18221
Name: count, dtype: int64

len(df)

df['substrate'] = df['substrate'].\
str.replace('r', 'R').\
str.replace('k', 'K').\
str.replace('n', 'N').\
str.replace('d', 'D').\
str.replace('h', 'H').\
str.replace('c', 'C').\
str.replace('X', '_').\
str.replace('p', 'P').\
str.replace('a', 'A').\
str.replace('v', 'V').\
str.replace('g', 'G').\
str.replace('m', 'M').\
str.replace('l', 'L').\
str.replace('q', 'Q').\
str.replace('f', 'F')

Split phosphosite sequence to single amino acid

df

	Kinase	gene_site	substrate	kinase_uniprot	kinase_paper	source
0	DYRK2	GLI2_S385	AEGLRPAsPLGLTQE	Q92630	DYRK2	pplus
1	DYRK2	SIAH2_S68	GGGAGPVsPQHHELT	Q92630	DYRK2	pplus
2	DYRK2	CARHSP1_S41	LRGNVVPsPLPtRRt	Q92630	DYRK2	pplus
3	DYRK2	DCX_S306	GPMRRSKsPADSANG	Q92630	DYRK2	pplus
4	DYRK2	CDC25A_S283	PERsQEEsPPGSTKR	Q92630	DYRK2	pplus
...	...	...	...	...	...	...
216464	SPHK1	HERC2_S2928	AAVPFLAsDNEEEED	Q9NYA1	NaN	large_scale
216465	SPHK1	PARG_Y832	DALHFRRyLDQFVPE	Q9NYA1	NaN	large_scale
216466	SPHK1	G6PI_T109	ALRNRsNtPILVDGK	Q9NYA1	NaN	large_scale
216467	PRPK	TITIN_S20761	DAERKSWsTVTTECs	Q96S44	PRPK	large_scale
216468	PRPK	TITIN_S20768	sTVTTECsKTSFRVA	Q96S44	PRPK	large_scale

216469 rows × 6 columns

Split site sequence to individual columns

# Split the 'SITE_+/-7_AA' column into individual letters
split_cols = df['substrate'].apply(list).apply(pd.Series).fillna('_')

# Rename the new columns from -7 to 7
split_cols.columns = range(-7, 8)

# Concatenate the original dataframe with the split columns
df = pd.concat([df, split_cols], axis=1)

df

	Kinase	gene_site	substrate	kinase_uniprot	kinase_paper	source	-7	-6	-5	-4	...	-2	-1	0	1	2	3	4	5	6	7
0	DYRK2	GLI2_S385	AEGLRPAsPLGLTQE	Q92630	DYRK2	pplus	A	E	G	L	...	P	A	s	P	L	G	L	T	Q	E
1	DYRK2	SIAH2_S68	GGGAGPVsPQHHELT	Q92630	DYRK2	pplus	G	G	G	A	...	P	V	s	P	Q	H	H	E	L	T
2	DYRK2	CARHSP1_S41	LRGNVVPsPLPtRRt	Q92630	DYRK2	pplus	L	R	G	N	...	V	P	s	P	L	P	t	R	R	t
3	DYRK2	DCX_S306	GPMRRSKsPADSANG	Q92630	DYRK2	pplus	G	P	M	R	...	S	K	s	P	A	D	S	A	N	G
4	DYRK2	CDC25A_S283	PERsQEEsPPGSTKR	Q92630	DYRK2	pplus	P	E	R	s	...	E	E	s	P	P	G	S	T	K	R
...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...
216464	SPHK1	HERC2_S2928	AAVPFLAsDNEEEED	Q9NYA1	NaN	large_scale	A	A	V	P	...	L	A	s	D	N	E	E	E	E	D
216465	SPHK1	PARG_Y832	DALHFRRyLDQFVPE	Q9NYA1	NaN	large_scale	D	A	L	H	...	R	R	y	L	D	Q	F	V	P	E
216466	SPHK1	G6PI_T109	ALRNRsNtPILVDGK	Q9NYA1	NaN	large_scale	A	L	R	N	...	s	N	t	P	I	L	V	D	G	K
216467	PRPK	TITIN_S20761	DAERKSWsTVTTECs	Q96S44	PRPK	large_scale	D	A	E	R	...	S	W	s	T	V	T	T	E	C	s
216468	PRPK	TITIN_S20768	sTVTTECsKTSFRVA	Q96S44	PRPK	large_scale	s	T	V	T	...	E	C	s	K	T	S	F	R	V	A

216469 rows × 21 columns

Add extra columns

df['kinase'] = df['kinase_paper'].fillna(df['Kinase'])

df['on_tree']=df['kinase_paper'].notna().astype(int)

Save

Uncheck below

df.columns = df.columns.astype(str) # Parquet can only save df with string column name
df.to_parquet('raw/df.parquet')

Load data

Need to convert the number in the columns to integer type

df = pd.read_parquet('raw/df.parquet')
#Convert the number in the column name into integer
df.columns = [int(col) if col.lstrip('-').isdigit() else col for col in df.columns]

Visualize counts

df.kinase.value_counts().hist(bins=100)
plt.axvline(x=100, color='r', linestyle='--', linewidth=1)
plt.xlabel('count per kinase')
plt.ylabel('# kinase');

cnt = df.kinase.value_counts().reset_index()

cnt.columns = ['kinase','count']

cnt

	kinase	count
0	SRC	2605
1	EPHA3	2120
2	FES	2013
3	NTRK3	1927
4	ALK	1889
...	...	...
488	CAMK1G	1
489	CAMK2D iso8	1
490	PHKA1	1
491	VPRBP	1
492	NEK10	1

493 rows × 2 columns