In silico toxicity prediction of 1-phenyl-1-(quinazolin-4-yl) ethanol compounds by using Toxtree, pkCSM and preADMET

The 1-phenyl-1-(quinazolin-4-yl) ethanol compounds are alkaloids of quinozoline class found in many Hydrangeaceae families. A survey revealed that most of the identified quinazoline derivatives have anticancer activity. Toxicity prediction of 1-phenyl-1-(quinazolin-4-yl) ethanols compounds were performed to obtain the best three compounds with high activity and the lowest toxicity. Toxicity prediction was conducted using Toxtree, pkCSM and PreADMET. The 2D structure of compounds were formed using ChemDraw. The decision tree approach was used in Toxtree application with endpoints including Cramer rules, Kroes TTC, carcinogenicity (genotoxic and non genotoxic) and in vitro mutagenicity. Graph based signature was used in pkCSM application with endpoints including mutagenicity, maximum daily dose, LD50 and hepatotoxicity. In PreADMET application, a method based on drugs similarity and ADMET properties was used with endpoints including mutagenicity, carcinogenicity to rat and mice. The results of data analysis showed that the best three anticancer compounds that have high activity and the lowest toxicity are compounds 14, 16 and 19.


Methods
The 2D structures of 44 compounds of 1-phenyl-1-(quinazolin-4-yl) ethanol were prepared using ChemDraw 2016. The 1-phenyl-1-(quinazoline-4-yl) ethanol compounds were screened using pkCSM to find out whether the compounds conform the Lipinski's rule of Five. The unconform compounds maximum 2 endpoints of Lipinski's rule of Five were eliminated. The toxicity of screened 1-phenyl-1-(quinazolin-4-yl) ethanol compound were predicted using Toxtree, pkCSM and PreADMET. The endpoints selected in Toxtree were Cramer rule, Kroes TTC decision tree, carcinogenicity (genotox and non genotox) mutagenicity rule base by ISS, and in vitro mutagenicity (Ames test) alerts by ISS. Open Babel GUI is used in pkCSM to create a compound SMILE format. The selected endpoint in pkCSM were Ames Toxicity, Maximum Tolerated Dose, Rat Acute Oral Toxicity (LD 50 ) and hepatotoxicity. The selected endpoints in PreADMET were Ames Test and Rodent Carcinogenicity (Mice and Rat).

Data Analysis
The toxicity prediction results of 1-phenyl-1-(quinazolin-4-yl) ethanol compounds were quantitative and qualitative data. Qualitative data were expressed in positive and negative statements. Then made in the form of scoring, where the positive toxic score was 1 and negative toxic scored was 2. The data analysis used the scoring model by summing all endpoints of Toxtree, pkCSM and PreADMET to obtain five compounds with the lowest toxicity effect (largest score). Then five of 1-phenyl-1-(quinazolin-4-yl) ethanol compounds which have high activity based on in vitro test of Kuroiwa et al. (2015) against cell line A549 (lung) were selected. The best compound is obtained through the selected scoring model by comparing any compounds having a low toxic effect and followed by the most amount of toxic negative endpoints. The next step for getting the three compounds that have the highest activity with the lowest toxicity was comparing the highest scores and the smallest IC 50 values among the five compounds.

RESULTS AND DISCUSSION Lipinski's rule of five calculations
The Lipinski's rule of five calculations were performed to determine the degree of absorption or permeability of compounds against lipid bilayers in the human body. The Lipinski rule is a parameter that demonstrates the oral bioavailability of a compound. Good bioavailability will satisfy the Lipinski rule where the maximum molecular weight of the compound is 500, the log P is not greater than 5, the hydrogen bond donor is less than 5, and hydrogen bond acceptor is less than 10 (Lipinski et al., 2012). The results of Lipinski's rule of Five calculations using pkCSM are presented in Table II. Based on results of Lipinski's Rule of Five calculations, all of 1-phenyl-1-(quinazolin-4-yl) ethanol compounds conform Lipinski's rule. All compounds were predicted having good absorptivity for an oral medication (Wulandari and Kristin, 2010). Based on research conducted by Veber et al. (2002) concluded that a compound with lower molecular weight, log P, hydrogen bond donor, and hydrogen bond acceptor has the higher bioavailability.
Compound 6 has the highest value of LD 50 endpoint of 2.996 mol/kg. Compound 4f has the highest value at the maximum daily dose endpoint of 4.188 mg/kg/day. The higher maximum daily dose and LD 50 value in the acute toxicity test of the compound, the compound will not have toxic effect on the mice. Based on the mutagenicity endpoint of pKCSM, the compounds 1b, 1c, 1f, 1m, 4f, 6, 31, 32, 33 and 34 are mutagenic (score 1) whereas the other compounds are non-mutagenic (score 2). The last parameter of pkCSM is hepatotoxic. There are 22 hepatotoxic compounds (score 1) and 22 non-hepatotoxic compounds (score 2).
At the Ames test endpoint of PreADMET, there are 14 mutagenic compounds (score 1) and 30 other compounds are non-mutagenic compounds (score 2). The positive test results on Ames test indicate that the compound is mutagenic and has the possibility as carcinogenic. In the prediction of carcinogenicity in rat produced 11 carcinogenic positive compounds (score 1) and 33 other compounds are negative carcinogenic (score 2). While in the prediction of carcinogenicity in mice, all of compounds are not carcinogenicity (score 2).
In the study of Kuroiwa et al. (2015) obtained compounds 14, 16, 17, 19 and 31 which have the best activity with IC 50 0.053 μM, 0.038 μM, 0.027 μM, 0.058 μM and 0.035 μM on cell line A549 (lung). The results of the toxicity prediction showed the five compounds having the lowest toxicity (largest scores), i.e. compounds 1j, 5, 14, 16 and 19. Compound 31 was not selected because of hepatotoxic, non-genotoxic and mutagenic carcinogens. While the compound 17 was not selected because of hepatotoxic. Compounds 1j, 5, 14, 16 and 19 are negative genotoxic carcinogens and non-genotoxic carcinogens, in vitro mutagenicity (Ames test), hepatotoxicity, and carcinogenicity in mice and rat. The highest maximum daily dose value among the five compounds is compound 1j, that is 3.034 mg/kg/day while the highest LD 50 value is compound 16, that is 2.642 mol/kg.
Based on in vitro test results Kuroiwa et al. (2015) compounds 1j, 5, 14, 16 and 19 have IC 50 values of 20 μM, 1.8 μM, 0.053 μM, 0.038 μM and 0.058 μM. The compound with smaller IC 50 is a compound that has higher activity as anticancer. Compounds 1j and 5 are not selected because they have a lower activity value compared to the other three compounds. Compounds that have small IC 50 values with low toxicity effects, i.e. compounds 14, 16 and 19.