Compound Multi-Dimensional Scaling (MDS) Visualisation using ChEMBL, Pandas, RDKit, SciPy, SciKit and ggplot¶

George Papadatos, ChEMBL Group.¶

In [160]:

import requests

import pandas as pd

from rdkit.Chem.Draw import IPythonConsole
from rdkit.Chem import PandasTools
from rdkit.Chem import AllChem as Chem
from rdkit.Chem import DataStructs

from scipy.spatial.distance import *
import numpy as np
from sklearn import manifold

from ggplot import *

1) Fetch and process data from the ChEMBL web services¶

In [161]:

TRGT = 'CHEMBL5024'

In [162]:

re = requests.get('https://www.ebi.ac.uk/chemblws/targets/{0}/bioactivities.json'.format(TRGT))

In [163]:

data = pd.DataFrame(re.json()['bioactivities'])

In [164]:

data.head()

Out[164]:

	activity_comment	assay_chemblid	assay_description	assay_type	bioactivity_type	ingredient_cmpd_chemblid	name_in_reference	operator	organism	parent_cmpd_chemblid	reference	target_chemblid	target_confidence	target_name	units	value
0	Unspecified	CHEMBL2330631	Inhibition of ATR-mediated CHK1 phosphorylation at serine 345 in human HT29 cells after 1 hr in presence of 4-nitroquinoline 1-oxide	B	IC50	CHEMBL2325438	36	=	Homo sapiens	CHEMBL2325438	J. Med. Chem., (2013) 56:5:2125	CHEMBL5024	9	Serine-protein kinase ATR	nM	9200
1	Unspecified	CHEMBL2330631	Inhibition of ATR-mediated CHK1 phosphorylation at serine 345 in human HT29 cells after 1 hr in presence of 4-nitroquinoline 1-oxide	B	IC50	CHEMBL2325726	35	=	Homo sapiens	CHEMBL2325726	J. Med. Chem., (2013) 56:5:2125	CHEMBL5024	9	Serine-protein kinase ATR	nM	15000
2	Unspecified	CHEMBL2330631	Inhibition of ATR-mediated CHK1 phosphorylation at serine 345 in human HT29 cells after 1 hr in presence of 4-nitroquinoline 1-oxide	B	IC50	CHEMBL2325723	32	=	Homo sapiens	CHEMBL2325723	J. Med. Chem., (2013) 56:5:2125	CHEMBL5024	9	Serine-protein kinase ATR	nM	7200
3	Unspecified	CHEMBL2330631	Inhibition of ATR-mediated CHK1 phosphorylation at serine 345 in human HT29 cells after 1 hr in presence of 4-nitroquinoline 1-oxide	B	IC50	CHEMBL2325714	22	>	Homo sapiens	CHEMBL2325714	J. Med. Chem., (2013) 56:5:2125	CHEMBL5024	9	Serine-protein kinase ATR	nM	30000
4	Unspecified	CHEMBL2330631	Inhibition of ATR-mediated CHK1 phosphorylation at serine 345 in human HT29 cells after 1 hr in presence of 4-nitroquinoline 1-oxide	B	IC50	CHEMBL2030442	5	=	Homo sapiens	CHEMBL2030442	J. Med. Chem., (2013) 56:5:2125	CHEMBL5024	9	Serine-protein kinase ATR	nM	1100

In [165]:

data = data.drop_duplicates(['parent_cmpd_chemblid'])

In [166]:

data.shape

Out[166]:

(135, 16)

Remove assays with less than 4 compounds¶

In [167]:

assays = data[['assay_chemblid','target_chemblid']]

In [168]:

assays = assays.groupby('assay_chemblid').count()

In [169]:

goodassays = list(assays.ix[assays.assay_chemblid >= 4].index)

In [170]:

data = data.ix[data.assay_chemblid.isin(goodassays)]

In [171]:

data.shape

Out[171]:

(124, 16)

2) Fetch SMILES and append to the frame - Convert to RDKit¶

In [172]:

def fetch_SMILES(chemblid):
    return str(requests.get('https://www.ebi.ac.uk/chemblws/compounds/{}.json'.format(chemblid)).json()['compound']['smiles'])

In [173]:

data['SMILES'] = data['parent_cmpd_chemblid'].map(fetch_SMILES)

In [174]:

PandasTools.AddMoleculeColumnToFrame(data, smilesCol = 'SMILES')

In [175]:

mols = data[['parent_cmpd_chemblid','name_in_reference','SMILES', 'ROMol', 'assay_chemblid']]

In [176]:

mols.shape

Out[176]:

(124, 5)

In [177]:

mols.head()

Out[177]:

	parent_cmpd_chemblid	name_in_reference	SMILES	assay_chemblid
0	CHEMBL2325438	36	CS(=O)(=O)C1(CC1)c2cc(nc(n2)c3cc(cc4[nH]ccc34)C(=O)N)N5CCOCC5	CHEMBL2330631
1	CHEMBL2325726	35	CNC(=O)c1cc(c2nc(cc(n2)C3(CC3)S(=O)(=O)C)N4CCOCC4)c5cc[nH]c5c1	CHEMBL2330631
2	CHEMBL2325723	32	CS(=O)(=O)C1(CC1)c2cc(nc(n2)c3cccc4[nH]c(cc34)C#N)N5CCOCC5	CHEMBL2330631
3	CHEMBL2325714	22	CS(=O)(=O)C1(CC1)c2cc(nc(n2)c3cccc(N)c3)N4CCOCC4	CHEMBL2330631
4	CHEMBL2030442	5	CS(=O)(=O)Cc1cc(nc(n1)c2cccc3[nH]ccc23)N4CCOCC4	CHEMBL2330631

3) Calculate Fingerprint Tanimoto Distance Matrix (efficiently thanks to scipy)¶

In [178]:

fps = [Chem.GetMorganFingerprintAsBitVect(m,2,nBits=2048) for m in mols['ROMol']]

In [179]:

dist_mat = squareform(pdist(fps,'jaccard'))

In [180]:

pd.DataFrame(dist_mat, columns = mols['parent_cmpd_chemblid'], index=mols['parent_cmpd_chemblid']).head()

Out[180]:

parent_cmpd_chemblid	CHEMBL2325438	CHEMBL2325726	CHEMBL2325723	CHEMBL2325714	CHEMBL2030442	CHEMBL199416	CHEMBL1095933	CHEMBL561907	CHEMBL371968	CHEMBL2325708	CHEMBL199050	CHEMBL1766788	CHEMBL1766794	CHEMBL2325715	CHEMBL1098243	CHEMBL1766790	CHEMBL200299	CHEMBL2325439	CHEMBL1766799	CHEMBL2325704	CHEMBL2325725	CHEMBL1766793	CHEMBL1766798	CHEMBL1766762	CHEMBL196764	CHEMBL563985	CHEMBL2325705	CHEMBL198740	CHEMBL1766792	CHEMBL1766774	CHEMBL1766802	CHEMBL2325712	CHEMBL2325707	CHEMBL1766784	CHEMBL113	CHEMBL1766768	CHEMBL2325710	CHEMBL2325440	CHEMBL1766791	CHEMBL2325711	CHEMBL1766769	CHEMBL1766780	CHEMBL1765107	CHEMBL1766803	CHEMBL2325720	CHEMBL1766777	CHEMBL2325713	CHEMBL1766781	CHEMBL1766787	CHEMBL1766796	CHEMBL1766770	CHEMBL1766804	CHEMBL2325698	CHEMBL197227	CHEMBL2325709	CHEMBL1766800	CHEMBL196759	CHEMBL1940990	CHEMBL1766760	CHEMBL1766795	CHEMBL2325721	CHEMBL2321916	CHEMBL1766759	CHEMBL1766764	CHEMBL1766778	CHEMBL2325703	CHEMBL2325724	CHEMBL1766786	CHEMBL562999	CHEMBL199051	CHEMBL436767	CHEMBL200275	CHEMBL197439	CHEMBL1766789	CHEMBL369969	CHEMBL1766766	CHEMBL1095627	CHEMBL1095626	CHEMBL1094981	CHEMBL1766773	CHEMBL1766765	CHEMBL2325719	CHEMBL1098242	CHEMBL1766801	CHEMBL1766763	CHEMBL1940980	CHEMBL2325699	CHEMBL2325702	CHEMBL2030436	CHEMBL1766785	CHEMBL2325717	CHEMBL1766775	CHEMBL1766761	CHEMBL1766771	CHEMBL2325696	CHEMBL2325700	CHEMBL1766767	CHEMBL196964	CHEMBL2325722	CHEMBL1766782	CHEMBL1766783	CHEMBL1940981	CHEMBL199216	CHEMBL1766772	CHEMBL370108	CHEMBL198801	CHEMBL222102	CHEMBL2325695	CHEMBL381025	CHEMBL197239	CHEMBL1940987	CHEMBL2325716	CHEMBL1095932	CHEMBL199424	CHEMBL2325706	CHEMBL2325697	CHEMBL197219	CHEMBL1766776	CHEMBL382501	CHEMBL1097942	CHEMBL2325701	CHEMBL1766797	CHEMBL2325718	CHEMBL1766779
parent_cmpd_chemblid
CHEMBL2325438	0.000000	0.181818	0.447368	0.457143	0.500000	0.808989	0.786408	0.801887	0.813953	0.388889	0.788235	0.861702	0.863158	0.459459	0.790476	0.877551	0.795181	0.230769	0.848485	0.261538	0.181818	0.872340	0.868687	0.888889	0.811765	0.790000	0.298507	0.795455	0.858696	0.877778	0.869565	0.441558	0.328571	0.850575	0.923077	0.873563	0.441558	0.230769	0.881720	0.410959	0.879121	0.898876	0.900000	0.880435	0.218750	0.866667	0.439394	0.891304	0.881720	0.872340	0.887640	0.844444	0.448718	0.813187	0.418919	0.847826	0.808989	0.904348	0.900000	0.877551	0.347826	0.424658	0.894118	0.898876	0.903226	0.506667	0.264706	0.861702	0.794118	0.811765	0.813953	0.790698	0.804878	0.863158	0.716049	0.881720	0.762887	0.765957	0.809091	0.835165	0.891304	0.323529	0.798165	0.857143	0.880435	0.893805	0.453333	0.646341	0.584416	0.879121	0.394366	0.870968	0.891566	0.875000	0.418919	0.578947	0.842697	0.811765	0.347222	0.900000	0.875000	0.913043	0.813187	0.882979	0.755814	0.815217	0.800000	0.253731	0.808989	0.808511	0.884956	0.416667	0.796117	0.787234	0.328571	0.418919	0.817204	0.879121	0.826087	0.801887	0.589744	0.881188	0.416667	0.855556
CHEMBL2325726	0.181818	0.000000	0.461538	0.493151	0.513158	0.813187	0.754902	0.794393	0.818182	0.405405	0.793103	0.852632	0.854167	0.473684	0.783019	0.868687	0.800000	0.253731	0.840000	0.283582	0.205882	0.863158	0.860000	0.879121	0.816092	0.757576	0.318841	0.800000	0.849462	0.868132	0.884211	0.455696	0.347222	0.879121	0.925000	0.863636	0.455696	0.253731	0.872340	0.426667	0.869565	0.888889	0.890110	0.882979	0.242424	0.857143	0.455882	0.881720	0.872340	0.863158	0.877778	0.860215	0.462500	0.817204	0.434211	0.838710	0.813187	0.905983	0.888889	0.868687	0.342857	0.440000	0.883721	0.888889	0.893617	0.519481	0.285714	0.852632	0.786408	0.816092	0.818182	0.795455	0.809524	0.854167	0.722892	0.872340	0.767677	0.770833	0.801802	0.826087	0.881720	0.366197	0.790909	0.872340	0.870968	0.895652	0.467532	0.654762	0.594937	0.818182	0.432432	0.861702	0.880952	0.865169	0.434211	0.589744	0.833333	0.816092	0.428571	0.890110	0.865169	0.914530	0.817204	0.873684	0.775281	0.819149	0.804598	0.275362	0.813187	0.824742	0.886957	0.432432	0.764706	0.778947	0.347222	0.434211	0.821053	0.869565	0.829787	0.794393	0.600000	0.872549	0.432432	0.846154
CHEMBL2325723	0.447368	0.461538	0.000000	0.434783	0.520000	0.795455	0.809524	0.824074	0.800000	0.410959	0.743902	0.885417	0.886598	0.394366	0.813084	0.900000	0.780488	0.424658	0.871287	0.287879	0.461538	0.895833	0.901961	0.913043	0.811765	0.813725	0.323529	0.781609	0.882979	0.902174	0.869565	0.461538	0.352113	0.913043	0.923077	0.886364	0.461538	0.424658	0.905263	0.432432	0.891304	0.923077	0.923913	0.880435	0.416667	0.891304	0.415385	0.914894	0.905263	0.895833	0.900000	0.818182	0.468354	0.786517	0.440000	0.884211	0.767442	0.864865	0.926829	0.900000	0.416667	0.357143	0.919540	0.923077	0.879121	0.526316	0.361111	0.885417	0.817308	0.811765	0.813953	0.776471	0.804878	0.886598	0.776471	0.893617	0.800000	0.804124	0.830357	0.860215	0.840909	0.459459	0.819820	0.831461	0.892473	0.913043	0.473684	0.576923	0.654321	0.903226	0.347826	0.894737	0.904762	0.887640	0.440000	0.650000	0.855556	0.797619	0.275362	0.923913	0.912088	0.873874	0.772727	0.858696	0.811111	0.775281	0.771084	0.440000	0.795455	0.782609	0.904348	0.298507	0.819048	0.787234	0.352113	0.440000	0.777778	0.903226	0.813187	0.824074	0.589744	0.902913	0.218750	0.880435
CHEMBL2325714	0.457143	0.493151	0.434783	0.000000	0.597222	0.784810	0.776596	0.793814	0.773333	0.507042	0.743243	0.831325	0.833333	0.400000	0.781250	0.863636	0.750000	0.455882	0.818182	0.390625	0.493151	0.843373	0.853933	0.875000	0.786667	0.780220	0.424242	0.753247	0.827160	0.833333	0.810127	0.533333	0.449275	0.860759	0.911765	0.871795	0.552632	0.455882	0.853659	0.527778	0.878049	0.871795	0.858974	0.822785	0.447761	0.820513	0.211538	0.864198	0.853659	0.843373	0.887500	0.810127	0.558442	0.790123	0.534247	0.829268	0.769231	0.884615	0.885714	0.850575	0.447761	0.409091	0.849315	0.871795	0.878049	0.621622	0.478873	0.831325	0.784946	0.802632	0.789474	0.763158	0.777778	0.833333	0.779221	0.880952	0.720930	0.722892	0.801980	0.784810	0.891566	0.424242	0.790000	0.825000	0.865854	0.894231	0.567568	0.597222	0.565217	0.850000	0.400000	0.827160	0.876712	0.873418	0.534247	0.558824	0.837500	0.770270	0.420290	0.873418	0.858974	0.894231	0.775000	0.841463	0.723684	0.777778	0.773333	0.471429	0.769231	0.785714	0.884615	0.400000	0.787234	0.790698	0.449275	0.534247	0.765432	0.835443	0.790123	0.793814	0.591549	0.855556	0.400000	0.822785
CHEMBL2030442	0.500000	0.513158	0.520000	0.597222	0.000000	0.777778	0.808081	0.800000	0.782051	0.373134	0.720000	0.850575	0.852273	0.573333	0.811881	0.868132	0.760000	0.478873	0.836957	0.317460	0.513158	0.862069	0.858696	0.892857	0.794872	0.800000	0.328125	0.762500	0.847059	0.853659	0.845238	0.430556	0.358209	0.879518	0.915493	0.848101	0.430556	0.478873	0.885057	0.397059	0.855422	0.890244	0.891566	0.857143	0.471429	0.841463	0.608696	0.882353	0.872093	0.862069	0.864198	0.802469	0.480000	0.737500	0.405797	0.848837	0.746835	0.866667	0.905405	0.855556	0.513889	0.520548	0.884615	0.890244	0.895349	0.306452	0.500000	0.850575	0.804124	0.794872	0.797468	0.756410	0.786667	0.852273	0.689189	0.858824	0.784946	0.775281	0.807692	0.807229	0.869048	0.513889	0.807692	0.817073	0.870588	0.886792	0.506849	0.322581	0.327869	0.869048	0.513889	0.847059	0.896104	0.850000	0.472222	0.316667	0.814815	0.779221	0.526316	0.891566	0.878049	0.886792	0.753086	0.860465	0.765432	0.756098	0.733333	0.493151	0.777778	0.764706	0.866667	0.492958	0.806122	0.770115	0.358209	0.472222	0.728395	0.855422	0.783133	0.811881	0.254237	0.860215	0.492958	0.843373

4) Perform MDS with SciKit Learn¶

In [181]:

mds = manifold.MDS(n_components=2, dissimilarity="precomputed", random_state=3, n_jobs = 2)

In [182]:

results = mds.fit(dist_mat)

In [183]:

coords = results.embedding_

Append coordinates to frame¶

In [184]:

mols['X'] = [c[0] for c in coords]

In [185]:

mols['Y'] = [c[1] for c in coords]

In [186]:

mols.head()

Out[186]:

	parent_cmpd_chemblid	name_in_reference	SMILES	assay_chemblid	X	Y
0	CHEMBL2325438	36	CS(=O)(=O)C1(CC1)c2cc(nc(n2)c3cc(cc4[nH]ccc34)C(=O)N)N5CCOCC5	CHEMBL2330631	-0.063416	-0.440131
1	CHEMBL2325726	35	CNC(=O)c1cc(c2nc(cc(n2)C3(CC3)S(=O)(=O)C)N4CCOCC4)c5cc[nH]c5c1	CHEMBL2330631	-0.095068	-0.423555
2	CHEMBL2325723	32	CS(=O)(=O)C1(CC1)c2cc(nc(n2)c3cccc4[nH]c(cc34)C#N)N5CCOCC5	CHEMBL2330631	0.050322	-0.574499
3	CHEMBL2325714	22	CS(=O)(=O)C1(CC1)c2cc(nc(n2)c3cccc(N)c3)N4CCOCC4	CHEMBL2330631	0.017721	-0.280538
4	CHEMBL2030442	5	CS(=O)(=O)Cc1cc(nc(n1)c2cccc3[nH]ccc23)N4CCOCC4	CHEMBL2330631	-0.129206	-0.343458

4) Plot MDS Map¶

In [187]:

rcParams['figure.figsize'] = 12,12

In [188]:

scatter(mols['X'], mols['Y'])

Out[188]:

<matplotlib.collections.PathCollection at 0x83f6410>

Sanity check: Do assay_ids correspond with regions of chemical space? Hint: Yes.¶

In [189]:

ggplot(aes(x='X', y='Y', colour='assay_chemblid'), data=mols) + geom_point() 

Out[189]:

<ggplot: (8979761)>

In [190]:

#TODO: D3 interactive visualisation