would you be interested in an implementation of "Positional Analogue Scanning"? [feature request]

[UnixJunkie]

Description:

There is a recent paper in J. Med. Chem.:
Positional Analogue Scanning: An Effective Strategy for Multiparameter Optimization in Drug Design
https://pubs.acs.org/doi/10.1021/acs.jmedchem.9b02092

Maybe I can provide an implementation.

[UnixJunkie]

I suspect that even for purely computational studies, it could be interesting/useful.
It could go as a script in Contrib/ at first.

[greglandrum]

[Meta] I moved this into Discussions because I am trying to stop using the Issue tracker for questions and discussions.

[greglandrum]

@UnixJunkie yeah, I saw that paper and definitely think it would be interesting and useful functionality to have in the RDKit.

It probably does make sense to start with it in Contrib in order to get a sense of what the API should be and how it's best used. Then we can think about moving it into the core RDKit later if it looks like that makes sense.

[UnixJunkie]

Ok, I'll try to program something then.

[UnixJunkie]

I end up with that:

#!/usr/bin/env python3

# "Positional Analogue Scanning: An Effective Strategy for
# Multiparameter Optimization in Drug Design".
# Pennington, L. D., Aquila, B. M., Choi, Y., Valiulin, R. A., & Muegge, I.
# Journal of medicinal chemistry (2020).
# https://doi.org/10.1021/acs.jmedchem.9b02092

import rdkit
from rdkit import Chem
from rdkit.Chem import AllChem

def positional_analog_scan(mol, smarts_patt = '[cH]',
                           smi_substs = ['N','CF','CC','CO',
                                         'CCN','CCl','CC(F)(F)(F)','COC']):
    res = []
    patt = Chem.MolFromSmarts(smarts_patt)
    for smi in smi_substs:
        subst = Chem.MolFromSmiles(smi)
        analogs = AllChem.ReplaceSubstructs(mol, patt, subst)
        for a in analogs:
            analog_smi = Chem.MolToSmiles(a) # canonicalization
            if not analog_smi in res: # remove duplicates
                res.append(analog_smi)
    return res

# test
chinine_smi = \
'[H][C@@]1([C@@H](C2=CC=NC3=CC=C(C=C23)OC)O)C[C@@H]4CC[N@]1C[C@@H]4C=C'
print('%s\tchinine' % chinine_smi)
mol = Chem.MolFromSmiles(chinine_smi)
analogs = positional_analog_scan(mol)
for i, a in enumerate(analogs):
    print('%s\tchinine_%02d' % (a, i))

[UnixJunkie]

It looks like it is working correctly (by looking at the corresponding molecules in 2D):

[H][C@@]1([C@@H](C2=CC=NC3=CC=C(C=C23)OC)O)C[C@@H]4CC[N@]1C[C@@H]4C=C   chinine
C=C[C@H]1CN2CC[C@H]1C[C@H]2[C@H](O)c1ncnc2ccc(OC)cc12   chinine_00
C=C[C@H]1CN2CC[C@H]1C[C@H]2[C@H](O)c1cnnc2ccc(OC)cc12   chinine_01
C=C[C@H]1CN2CC[C@H]1C[C@H]2[C@H](O)c1ccnc2ncc(OC)cc12   chinine_02
C=C[C@H]1CN2CC[C@H]1C[C@H]2[C@H](O)c1ccnc2cnc(OC)cc12   chinine_03
C=C[C@H]1CN2CC[C@H]1C[C@H]2[C@H](O)c1ccnc2ccc(OC)nc12   chinine_04
C=C[C@H]1CN2CC[C@H]1C[C@H]2[C@H](O)c1c(F)cnc2ccc(OC)cc12        chinine_05
C=C[C@H]1CN2CC[C@H]1C[C@H]2[C@H](O)c1cc(F)nc2ccc(OC)cc12        chinine_06
C=C[C@H]1CN2CC[C@H]1C[C@H]2[C@H](O)c1ccnc2c(F)cc(OC)cc12        chinine_07
C=C[C@H]1CN2CC[C@H]1C[C@H]2[C@H](O)c1ccnc2cc(F)c(OC)cc12        chinine_08
C=C[C@H]1CN2CC[C@H]1C[C@H]2[C@H](O)c1ccnc2ccc(OC)c(F)c12        chinine_09
C=C[C@H]1CN2CC[C@H]1C[C@H]2[C@H](O)c1c(C)cnc2ccc(OC)cc12        chinine_10
C=C[C@H]1CN2CC[C@H]1C[C@H]2[C@H](O)c1cc(C)nc2ccc(OC)cc12        chinine_11
C=C[C@H]1CN2CC[C@H]1C[C@H]2[C@H](O)c1ccnc2c(C)cc(OC)cc12        chinine_12
C=C[C@H]1CN2CC[C@H]1C[C@H]2[C@H](O)c1ccnc2cc(C)c(OC)cc12        chinine_13
C=C[C@H]1CN2CC[C@H]1C[C@H]2[C@H](O)c1ccnc2ccc(OC)c(C)c12        chinine_14
C=C[C@H]1CN2CC[C@H]1C[C@H]2[C@H](O)c1c(O)cnc2ccc(OC)cc12        chinine_15
C=C[C@H]1CN2CC[C@H]1C[C@H]2[C@H](O)c1cc(O)nc2ccc(OC)cc12        chinine_16
C=C[C@H]1CN2CC[C@H]1C[C@H]2[C@H](O)c1ccnc2c(O)cc(OC)cc12        chinine_17
C=C[C@H]1CN2CC[C@H]1C[C@H]2[C@H](O)c1ccnc2cc(O)c(OC)cc12        chinine_18
C=C[C@H]1CN2CC[C@H]1C[C@H]2[C@H](O)c1ccnc2ccc(OC)c(O)c12        chinine_19
C=C[C@H]1CN2CC[C@H]1C[C@H]2[C@H](O)c1c(CN)cnc2ccc(OC)cc12       chinine_20
C=C[C@H]1CN2CC[C@H]1C[C@H]2[C@H](O)c1cc(CN)nc2ccc(OC)cc12       chinine_21
C=C[C@H]1CN2CC[C@H]1C[C@H]2[C@H](O)c1ccnc2c(CN)cc(OC)cc12       chinine_22
C=C[C@H]1CN2CC[C@H]1C[C@H]2[C@H](O)c1ccnc2cc(CN)c(OC)cc12       chinine_23
C=C[C@H]1CN2CC[C@H]1C[C@H]2[C@H](O)c1ccnc2ccc(OC)c(CN)c12       chinine_24
C=C[C@H]1CN2CC[C@H]1C[C@H]2[C@H](O)c1c(Cl)cnc2ccc(OC)cc12       chinine_25
C=C[C@H]1CN2CC[C@H]1C[C@H]2[C@H](O)c1cc(Cl)nc2ccc(OC)cc12       chinine_26
C=C[C@H]1CN2CC[C@H]1C[C@H]2[C@H](O)c1ccnc2c(Cl)cc(OC)cc12       chinine_27
C=C[C@H]1CN2CC[C@H]1C[C@H]2[C@H](O)c1ccnc2cc(Cl)c(OC)cc12       chinine_28
C=C[C@H]1CN2CC[C@H]1C[C@H]2[C@H](O)c1ccnc2ccc(OC)c(Cl)c12       chinine_29
C=C[C@H]1CN2CC[C@H]1C[C@H]2[C@H](O)c1c(C(F)(F)F)cnc2ccc(OC)cc12 chinine_30
C=C[C@H]1CN2CC[C@H]1C[C@H]2[C@H](O)c1cc(C(F)(F)F)nc2ccc(OC)cc12 chinine_31
C=C[C@H]1CN2CC[C@H]1C[C@H]2[C@H](O)c1ccnc2c(C(F)(F)F)cc(OC)cc12 chinine_32
C=C[C@H]1CN2CC[C@H]1C[C@H]2[C@H](O)c1ccnc2cc(C(F)(F)F)c(OC)cc12 chinine_33
C=C[C@H]1CN2CC[C@H]1C[C@H]2[C@H](O)c1ccnc2ccc(OC)c(C(F)(F)F)c12 chinine_34
C=C[C@H]1CN2CC[C@H]1C[C@H]2[C@H](O)c1c(OC)cnc2ccc(OC)cc12       chinine_35
C=C[C@H]1CN2CC[C@H]1C[C@H]2[C@H](O)c1cc(OC)nc2ccc(OC)cc12       chinine_36
C=C[C@H]1CN2CC[C@H]1C[C@H]2[C@H](O)c1ccnc2c(OC)cc(OC)cc12       chinine_37
C=C[C@H]1CN2CC[C@H]1C[C@H]2[C@H](O)c1ccnc2cc(OC)c(OC)cc12       chinine_38
C=C[C@H]1CN2CC[C@H]1C[C@H]2[C@H](O)c1ccnc2ccc(OC)c(OC)c12       chinine_39

[UnixJunkie]

@greglandrum should I transform this into a PR?

[greglandrum]

Sorry, I'm behind and haven't had a chance to look at this. Will try to do it in the next day or so.

[UnixJunkie]

no problem, not urgent