Gnina rescore with docked ligand

Setup

from kdock.data.core import Data
from kdock.gnina.dock import *
from kdock.gnina.rescore import *
import pandas as pd, numpy as np

# setup_gnina_local('v1.1') # <= v1.1 allows cpu environment

Reading package lists...
Building dependency tree...
Reading state information...
openbabel is already the newest version (3.0.0+dfsg-3ubuntu3).
0 upgraded, 0 newly installed, 0 to remove and 6 not upgraded.
Downloading v1.1 gnina
Finish setup!

# setup_gnina_docker()

Pulling GNINA Docker image: gnina/gnina
GNINA Docker image is ready.

Rescore

Get folder that contains the AF3 .cif files

local ./gnina

df = get_gnina_rescore_folder('~/mirati_structure',
                              rec_chain_id='A',
                              lig_chain_id='L',
                             )

CPU times: user 1.19 s, sys: 257 ms, total: 1.44 s
Wall time: 21min 55s

if vinardo score only:

df = get_gnina_rescore_folder('~/mirati_structure',
                              rec_chain_id='A',
                              lig_chain_id='L',
                              CNN_affinity=False,
                              vinardo=True,
                             )

Docker gnina (latest)

df = get_gnina_rescore_folder('~/mirati_structure',
                              rec_chain_id='A',
                              lig_chain_id='L',
                              is_local=False)

Merge with target

df.ID = df.ID.str[:-15]

target = Data.get_mirati_g12d()

target.ID=target.ID.str.lower()

target['log_Kd'] = np.log(target['Kd'] + 1e-5)
target['log_IC50'] = np.log(target['IC50'] + 1e-5)
target['log_erk_IC50'] = np.log(target['erk_IC50'] + 1e-5)

out =target.merge(df)

out.head()

	ID	SMILES	Kd	IC50	erk_IC50	log_Kd	log_IC50	log_erk_IC50	binding_energy	uncertainty	RMSD	CNNscore	CNNaffinity	CNNvariance
0	us_1	CN1CCC[C@H]1COc1nc(N2CC3CCC(C2)N3)c2cnc(cc2n1)...	97.7	124.7	3159.1	4.581902	4.825911	8.058042	-13.02696	-0.30852	0.88857	0.67440	7.76933	0.23841
1	us_4	Oc1cc(-c2ncc3c(nc(OCCc4ccccn4)nc3c2F)N2CC3CCC(...	155.7	496.2	8530.0	5.047931	6.206979	9.051345	-11.44830	-0.70801	1.01023	0.67062	7.98086	0.29582
2	us_5	Cn1nccc1COc1nc(N2CC3CCC(C2)N3)c2cnc(c(F)c2n1)-...	294.8	722.9	8193.8	5.686297	6.583271	9.011133	-7.58041	-1.00321	0.91650	0.48476	7.27615	0.59758
3	us_6	Cc1cccnc1CCOc1nc(N2CC3CCC(C2)N3)c2cnc(c(F)c2n1...	442.2	434.1	11518.2	6.091762	6.073275	9.351684	-7.00049	-1.53471	4.55312	0.54943	6.14055	0.24137
4	us_7	Oc1cc(-c2ncc3c(nc(OCCc4ncccn4)nc3c2F)N2CC3CCC(...	463.5	1867.3	NaN	6.138806	7.532249	NaN	-13.12053	-0.93447	0.28536	0.71432	7.42845	0.04068

out.to_csv('mirati_gnina_v1_1.csv',index=False)

Merge with AF score

df=pd.read_csv('mirati_gnina_rescore_v1_3.csv')

af = pd.read_csv('AF3_mirati_660.csv')

cols = ['ID','chain_pair_pae_min_0_1',
       'chain_pair_pae_min_1_0', 'chain_pair_pae_min_1_1', 'chain_ptm_0',
       'chain_ptm_1', 'fraction_disordered', 'iptm', 'ptm', 'ranking_score',
       'iptm_ptm_add', 'chain_pair_pae_min_add', 'iptm_rnk', 'ptm_rnk',
       'iptm_ptm_rnk_add', 'chain_pair_pae_min_add_rnk', 'iptm_pae_add_rnk']

af = af[cols]

df = df.merge(af)

df.to_csv('mirati_gnina_v1_3_af.csv',index=False)

Correlation

out.corr(numeric_only=True)

	Kd	IC50	erk_IC50	log_Kd	log_IC50	log_erk_IC50	binding_energy	uncertainty	RMSD	CNNscore	CNNaffinity	CNNvariance
Kd	1.000000	0.137290	-0.088274	0.567310	0.081667	-0.167722	0.057558	-0.017178	-0.025133	-0.048843	-0.050384	0.025814
IC50	0.137290	1.000000	0.379993	0.409676	0.405385	0.242495	0.053759	0.043837	0.004114	-0.018452	-0.084494	-0.056643
erk_IC50	-0.088274	0.379993	1.000000	0.340125	0.712639	0.734366	0.021087	0.178889	-0.048670	0.003114	-0.090038	-0.111960
log_Kd	0.567310	0.409676	0.340125	1.000000	0.685839	0.489439	0.248571	-0.040234	0.127425	-0.239114	-0.297084	-0.032168
log_IC50	0.081667	0.405385	0.712639	0.685839	1.000000	0.782773	0.122564	0.225565	0.034728	-0.102558	-0.241604	-0.134573
log_erk_IC50	-0.167722	0.242495	0.734366	0.489439	0.782773	1.000000	0.104046	0.180925	-0.018172	-0.056561	-0.203590	-0.102189
binding_energy	0.057558	0.053759	0.021087	0.248571	0.122564	0.104046	1.000000	-0.446761	0.585579	-0.605980	-0.719846	0.146693
uncertainty	-0.017178	0.043837	0.178889	-0.040234	0.225565	0.180925	-0.446761	1.000000	-0.340409	0.415752	0.292460	-0.295152
RMSD	-0.025133	0.004114	-0.048670	0.127425	0.034728	-0.018172	0.585579	-0.340409	1.000000	-0.502661	-0.567085	0.241297
CNNscore	-0.048843	-0.018452	0.003114	-0.239114	-0.102558	-0.056561	-0.605980	0.415752	-0.502661	1.000000	0.804062	-0.475357
CNNaffinity	-0.050384	-0.084494	-0.090038	-0.297084	-0.241604	-0.203590	-0.719846	0.292460	-0.567085	0.804062	1.000000	-0.334174
CNNvariance	0.025814	-0.056643	-0.111960	-0.032168	-0.134573	-0.102189	0.146693	-0.295152	0.241297	-0.475357	-0.334174	1.000000

out.corr('spearman',numeric_only=True)

	Kd	IC50	erk_IC50	log_Kd	log_IC50	log_erk_IC50	binding_energy	uncertainty	RMSD	CNNscore	CNNaffinity	CNNvariance
Kd	1.000000	0.770684	0.484753	1.000000	0.770684	0.484753	0.307612	-0.041270	0.197978	-0.271837	-0.377634	-0.075319
IC50	0.770684	1.000000	0.821105	0.770684	1.000000	0.821105	0.214448	0.201271	0.024432	-0.102062	-0.317811	-0.171460
erk_IC50	0.484753	0.821105	1.000000	0.484753	0.821105	1.000000	0.148804	0.197881	-0.012566	-0.031280	-0.228966	-0.100639
log_Kd	1.000000	0.770684	0.484753	1.000000	0.770684	0.484753	0.307612	-0.041270	0.197978	-0.271837	-0.377634	-0.075319
log_IC50	0.770684	1.000000	0.821105	0.770684	1.000000	0.821105	0.214448	0.201271	0.024432	-0.102062	-0.317811	-0.171460
log_erk_IC50	0.484753	0.821105	1.000000	0.484753	0.821105	1.000000	0.148804	0.197881	-0.012566	-0.031280	-0.228966	-0.100639
binding_energy	0.307612	0.214448	0.148804	0.307612	0.214448	0.148804	1.000000	-0.318202	0.643279	-0.596055	-0.657687	0.110284
uncertainty	-0.041270	0.201271	0.197881	-0.041270	0.201271	0.197881	-0.318202	1.000000	-0.285817	0.383521	0.103260	-0.278986
RMSD	0.197978	0.024432	-0.012566	0.197978	0.024432	-0.012566	0.643279	-0.285817	1.000000	-0.608563	-0.558381	0.233253
CNNscore	-0.271837	-0.102062	-0.031280	-0.271837	-0.102062	-0.031280	-0.596055	0.383521	-0.608563	1.000000	0.765298	-0.483761
CNNaffinity	-0.377634	-0.317811	-0.228966	-0.377634	-0.317811	-0.228966	-0.657687	0.103260	-0.558381	0.765298	1.000000	-0.251215
CNNvariance	-0.075319	-0.171460	-0.100639	-0.075319	-0.171460	-0.100639	0.110284	-0.278986	0.233253	-0.483761	-0.251215	1.000000

out.corr('spearman',numeric_only=True)

	Kd	IC50	erk_IC50	binding_energy	uncertainty	RMSD	CNNscore	CNNaffinity	CNNvariance
Kd	1.000000	0.770684	0.484753	0.307595	-0.041586	0.198397	-0.318074	-0.343751	0.173958
IC50	0.770684	1.000000	0.821105	0.214860	0.201848	0.019858	-0.164505	-0.355147	0.205119
erk_IC50	0.484753	0.821105	1.000000	0.150561	0.201548	-0.023826	-0.118687	-0.287123	0.202999
binding_energy	0.307595	0.214860	0.150561	1.000000	-0.317127	0.642633	-0.723730	-0.723106	0.457403
uncertainty	-0.041586	0.201848	0.201548	-0.317127	1.000000	-0.290202	0.329470	0.111949	-0.021203
RMSD	0.198397	0.019858	-0.023826	0.642633	-0.290202	1.000000	-0.664650	-0.442909	0.441766
CNNscore	-0.318074	-0.164505	-0.118687	-0.723730	0.329470	-0.664650	1.000000	0.635484	-0.412043
CNNaffinity	-0.343751	-0.355147	-0.287123	-0.723106	0.111949	-0.442909	0.635484	1.000000	-0.269412
CNNvariance	0.173958	0.205119	0.202999	0.457403	-0.021203	0.441766	-0.412043	-0.269412	1.000000