Development of novel ALOX15 inhibitors combining dual machine learning filtering and fragment substitution optimisation approaches, molecular docking and dynamic simulation methods

Figures & data

Figure 1. Schematic diagram of the workflow of this study.

Figure 2. Internal/external validation ROC curves and confusion matrix for the Naive Bayesian model. (A)Confusion matrix for 10-fold validation of the model internal; (B) ROC curve for 10-fold validation of the model internal; (C) Confusion matrix for the internal external test of the model; (D) ROC curve for the external test of the model.

Table 1. Internal/external validation of the Naive Bayesian model.

Validation	Sensitivity	Specificity	ROC score	ROC rating	Precision	F-1 score	MCC
10-fold validation	0.92	0.86	0.86	Good	0.66	0.77	0.63
External test	0.90	0.77	0.92	Excellent	0.77	0.83	0.75

Figure 3. Confusion matrices for in-bag/out-of-bag validation/test set validation of the Recursive Partition model. (A) Confusion matrix for model in-bag validation; (B) Confusion matrix for model internal out-of-bag validation; (C) Confusion matrix for model test set validation.

Table 2. In-bag/out-of-bag validation of the Recursive Partition model.

Validation	Sensitivity	Specificity	AUC	Precision	F-1 score	MCC
In-bag training	0.93	0.90	0.97	0.83	0.87	0.81
Out-of-bag training	0.74	0.82	0.86	0.67	0.70	0.57

Table 3. Structure, Libdock and CDOCKER energy scores of the three lead molecules and the control inhibitor i472.

Molecule	Structure	Libdock score	CDOCKER ENERGY	CDOCKER INTERACTION ENERGY
OSI-930		110.36	−17.94	−12.13
SD-208		93.16	−23.89	−17.31
GS-444217		119.96	−33.87	−21.49
i472		119.59	−34.13	−29.64

Table 4. Novel molecules obtained by i472 fragment replacement optimisation.

Molecule	Structure	Libdock score	CDOCKER ENERGY	CDOCKER INTERACTION ENERGY	Number of favourable interactions	Average interaction distance
i472		119.59	−34.13	−29.64	14	3.59 Å
i472a		130.67	−44.09	−32.58	15	3.34 Å
i472b		125.91	−37.14	−31.08	15	3.37 Å
i472c		123.90	−32.02	−29.15	14	3.24 Å
i472d		122.04	−36.15	−33.88	13	3.45 Å
i472e		121.54	−35.06	−28.90	16	3.38 Å

Table 5. Structural and target interaction information for OSI-930 and its fragment replacement optimised molecule osia - osif.

Molecule	Structure	Libdock score	CDOCKER ENERGY	CDOCKER INTERACTION ENERGY	Number of favourable interactions	Average interaction distance
OSI-930		110.36	−17.94	−12.14	12	3.30 Å
osia		121.72	−20.40	−14.70	17	3.30 Å
osib		106.49	−15.74	−13.63	13	3.39 Å
osic		105.46	−19.97	−15.59	24	3.37 Å
osid		97.80	−15.04	−9.53	14	3.22 Å
osie		90.13	−13.32	−10.81	14	3.24 Å
osif		70.25	−14.59	−11.08	13	3.24 Å

Table 6. Structural and target interaction information for GS-444217 and its fragment substitution optimised molecule gsa–gsh.

Molecule	Structure	Libdock score	CDOCKER ENERGY	CDOCKER INTERACTION ENERGY	Number of favourable interactions	Average interaction distance
GS-444217		119.96	−33.87	−21.49	14	3.47 Å
gsa		148.47	−54.32	−29.64	22	3.42 Å
gsb		134.33	−34.37	−17.56	17	3.48 Å
gsc		129.87	−47.85	−24.82	20	3.43 Å
gsd		129.78	−31.20	−18.65	14	3.55 Å
gse		129.08	−35.76	−18.47	19	3.41 Å
gsf		125.36	−32.88	−27.89	23	3.48 Å
gsg		123.37	−33.89	−20.11	17	3.23 Å
gsh		120.52	−27.11	−19.99	17	4.21 Å

Figure 4. Structural superposition schematic of candidate compounds.

(A) Structural overlay of gsa/b/c/e, gsa is earthy yellow, gsb is red, gsc is purple, and gse is green; (B) Schematic representation of the binding of gsa/b/c/e to the active pocket of ALOX15 protein; (C) Structural overlay of i472a/b/e, i472a is cyan, i472 is orange, and i472e is grey; (D) Schematic representation of the binding of i472a/b/e to the active pocket of ALOX15 protein.

Figure 5. Final selection of the lead compound structure.

Table 7. ADME test results for the seven lead molecules.

Molecule	Molecule Weight (g/mol)	Log Po/w  (iLOGP)	Solubility	BBB Permeant	Number of Hydrogen Bond Acceptor	Number of Hydrogen Bond Donor Number	Number of rotatable bond	GI absorption	CYP2D6 inhibition	P-gp substrate
i472a	411.86	2.93	Soluble	No	4	3	10	High	Yes	Yes
i472b	412.88	3.61	Moderately soluble	Yes	3	2	9	High	Yes	No
i472e	408.88	3.54	Moderately soluble	Yes	3	3	6	High	Yes	No
gsa	416.45	3.35	Moderately soluble	Yes	6	0	7	High	Yes	Yes
gsb	397.48	2.83	Moderately soluble	Yes	4	1	7	High	Yes	Yes
gsc	413.48	3.30	Soluble	No	6	1	8	High	Yes	Yes
gse	411.46	3.32	Moderately soluble	Yes	6	0	7	High	Yes	Yes

Table 8. Results of teratogenicity, carcinogenicity to male/female Rat tests for seven lead molecules.

Molecule	Ames_Prediction	Ames_Probability	Ames_TOPKAT Score	Carcinogen_Prediction (Male/Female Rat)	Rat_Male_Carcinogen_ Probability	Rat_Female_Carcinogen_ Probability
i472a	Non-Mutagen	0.68	−2.59	Non-Carcinogen/Non-Carcinogen	0.33	0.19
i472b	Non-Mutagen	0.63	−4.04	Non-Carcinogen/Non-Carcinogen	0.32	0.18
i472e	Non-Mutagen	0.69	−2.25	Non-Carcinogen/Non-Carcinogen	0.31	0.18
gsa	Non-Mutagen	0.60	−5.03	Single-Carcinogen/Non-Carcinogen	0.42	0.22
gsb	Non-Mutagen	0.53	−6.74	Single-Carcinogen/Non-Carcinogen	0.46	0.23
gsc	Non-Mutagen	0.65	−3.54	Single-Carcinogen/Non-Carcinogen	0.47	0.25
gse	Non-Mutagen	0.63	−4.14	Single-Carcinogen/Non-Carcinogen	0.37	0.25

Figure 6. Molecular dynamics simulations of the seven candidate molecules in complex with ALOX15 protein. (A)fluctuations in the root mean square deviation of the ligand; (B) fluctuations in the root mean square fluctuation of the protein residues forming the complex; (C) fluctuations in the radius of gyration of the protein receptor; (D) fluctuations in the potential energy of the complex system.

Figure 7. Number of interactions between 7 candidate molecules and ALOX15 protein over time. (A) Number of interactions between gsa and protein over time; (B) Number of interactions between gsb and protein over time; (C) Number of interactions between gsc and protein over time; (D) Number of interactions between gse and protein over time; (E) Number of interactions between ia and protein over time; (F) Number of interactions between ib and protein over time; (G) Number of interactions between i.e. and protein over time. number of interactions over time; (G) number of interactions between i.e. and proteins over time.

Table 9. Binding energy of between the candidate compounds and ALOX15.

Energy	Molecule	Features
Binding energy(kJ/mol)	gsa	−156.065 ± 42.274
	gsb	−162.429 ± 43.326
	gsc	−158.314 ± 39.992
	gse	−153.042 ± 40.982
	ia	−202.140 ± 37.789
	ib	−205.482 ± 38.842
	ie	−204.671 ± 41.347

Data availability

The data that support this study are available from the corresponding author upon reasonable.