BIOINFORMATIC AND COMPUTATIONAL PHARMACOLOGY: NEW PERSPECTIVE IN DEPRESSION TREATMENT IN RISK FACTOR GROUP PATIENTS

Depression is an illness that affects mostly women, the treatment of this illness is represented by antidepressants and psychotherapy. Even if pregnancy is a wonderful time in a woman’s life, the number of women that suffer of depression is growing. Antidepressant medication in pregnancy is not recommended, because of negative effects on new-born. In this paper we use bioinformatic methods to ﬁ nd a new treatment of depression safe to use in pregnancy. Our results show that natural compounds resveratrol (8.67), melatonin (8.58) and linalyl acetate (9.40) have biological activity similar with antidepressants on serotonin transporter. Because majority of antidepressants have serotonin transporter as main target, biological activity on serotonin transporter may indicate that this natural compound could have antidepressant-like activity. Supplementary we made an Absorption, Distribution, Metabo-lism, Excretion, Toxicity in silico study, to complete the pharmacological pro ﬁ le of these natural compounds, in comparation with antidepressants sertraline and amitriptyline. We conclude that melatonin, resveratrol and linalyl acetate are tolerated well by the human body, and may represent a viable alternative in depression treatment, even for risk factor groups.


INTRODUCTION
Depression is an illness that affects women more than men (Albert, 2015). Up to 12% of pregnant women are affected by depression (Grote, 2010). The treatment for depression is generally represented by antidepressant medication and psychotherapy.
Utilization of antidepressant drugs in pregnancy could have negative effects on new-born like malformations. If the medication is taken in the fi rst trimester of gestation; antidepressants could cause a shorter duration of pregnancy, and changes in the foetus development (Yonkers, 2010). The newborns present smaller head circumference, lower Apgar score and lower weight, also the new-borns could present behaviour changes (Marcus, 2009;Wisner, 2009). This study's aim was to fi nd an alternative depression treatment, using natural compounds. In this study we use resveratrol, a polyphenol found in blueberries (Vaccinium myrtillus L.), melatonin an indole produced by pineal gland but also found in goji fruits (Lycium barbarum L. and Lycium chinense Mill.), and linalyl acetate (Lavandula angustifolia Mill.) an ester that is found in lavender essential oil.
In case of melatonin (Tamura, 2008) we fi nd that is safe to use in pregnancy. We fi nd that resveratrol is safe to use during pregnancy in case of rats (Madhyastha, 2013) we don't fi nd data of resveratrol administration on pregnant women. Unfortunately, we are unable to fi nd data regarding the safety of linalyl acetate in pregnancy.
Studies have linked melatonin (Srinivasan, 2009), resveratrol (Ge J-F, 2016) and linalyl acetate (Koulivand, 2013) effect on ameliorating depressive symptoms. The mechanism of action of those molecules is still unknown.
In this study we evaluated if the molecules have biological activity in serotonin transporter (SERT), the main target for clinically drugs used in treatment of depression.
We used predicted values of placental transfer index from Hewitt article (Hewitt, 2007). In our previous work (Udrea, unpublished) we predict the placental transfer index for antidepressants, antipsychotic drugs but also for resveratrol, melatonin and linalyl acetate by developing a QSAR equation (1).
Results indicate that the most of antidepressants and the natural compound cross the placental barrier, but most antipsychotics do not. Our results are correlated with the negative effects on foetus and new-born, caused by antidepressants (Marcus, 2009;Wisner, 2009).
In addition, we predict the in silico Absorption, Distribution, Metabolism, Excretion and Toxicity (ADMET) parameters for resveratrol, melatonin and linalyl acetate.
We compare the results with the predicted AD-MET values of clinically used drugs amitriptyline and sertraline. We have selected those two molecules because they have a similar biological activity with natural compounds on SERT.
ADMET analysis tries to reveal how a drug behaves in the human body. In this study we evaluated the Caco2 permeability, Central nervous system permeability, renal OCT2 substrate, CYP2D6 inhibition, CYP1A2 substrate, Maximum tolerated dose in humans, hERG I and hERG II inhibition.
(i) The Caco2 permeability refer to the ability of a compound to be absorbed.
(ii) The Central Nervous System (CNS) permeability to see if the compounds are distributed in the CNS.
(iii) The renal OCT2 substrate offers information about the possible contraindication regarding co-administrations, and if the compound could be metabolized by kidneys.
(iv) We have used the CYP2D6 and CYP1A2 interaction profi le to evaluate the safety of natural compounds regarding liver metabolization.
(v) Regarding the toxicity of the natural compounds we evaluated the maximum tolerated dose on humans and (vi) the human ether-a-go-go gene (hERG I and II) inhibition. The hERG I and II inhibition affect potassium ion channels that lead to development of long QT syndrome (Priest, 2008).
The biological activity K i (uM) on SERT was collected from PDSP K i database (PDSP) after we apply the logarithm function (2) to express the K i as pK i . (2)

Molecular modelling and minimum energy opti mizati on and descriptor calculati on
To fulfi ll our goal we obtained the 3D structure of clinically used antipsychotic drugs, resveratrol, melatonin, linalyl acetate and clinically used antidepressants in format .mol from CHEMBL data base (Team EBIW).
We minimized the molecules using Forcefi eld MMFF94x at a 0.05 gradient and after we applied Gasteiger partial charges. The minimization of the molecules was done using Molecular Operating Environment 10 (Molecular Operating Environment, 2010.10) software.
The power of prediction of our 3D-QSAR model is confi rmed by the statistical results (table1).
PLS regression was used in order to identify from a very large set of independent variables the dependent variables. The PLS parameters that we evaluated are: coeffi cient of correlation (R2), cross-validated coeffi cient of correlation (Q2), Standard deviation error of calibration (SDEC) and Standard deviation error of prediction (SDEP) presented in Table 1.

In silico ADMET
In silico ADMET (Absorption, Distribution, Metabolism, Excretion and Toxicity) is an important technique used in drug discovery. The ADMET parameters evaluated how a compound will act inside the human body. It is important to know the ADMET parameters before the administration of a new compound to avoid severe side effects.
In this article we have used pkCSM-pharmacokinetics database (pkCSM), where we introduced the SMILES coding of melatonin, resveratrol, linalyl acetate, amitriptyline and sertraline taken from PubChem database (PubChem).

QSAR results
The 3D-QSAR model equation was applied and led to proper statistical results (R2: 0.96; Q2: 0.89). The power of prediction of our 3D QSAR model is also validated by the best (0.002) and the worst residual values (0.51).
The prediction's accuracy is confi rmed by the low residual values; presented in table 2. Results point to similar pK i activity of the tested natural compounds and commonly used antidepressant drugs.
In the fi rst column we present the molecule name of the learning set compounds. In the second column we present experimental pK i values of antidepressants and antipsychotic drugs. Third column shows predicted values for antipsychotic, antidepressants melatonin, resveratrol and linalyl acetate. Last column displays residuals values between experimental pK i values and predicted pK i values. Antidepressant range on pK i is from 6.79 (trazodone) to 10.09 (paroxetine). The Natural compounds values vary from 8.59 (melatonin) to 9.40 (linalyl acetate). The predicted values of melatonin (8.58) and resveratrol (8.67) are similar to pK i value of amitriptyline, while linalyl acetate value of 9.40 is close to sertraline (9.58).
The similarities in values points to related effects on the serotonin transporter. In fi gure 1 we present the comparation between experimental pK i values and predicted pK i values of antidepressants and antipsychotic drugs from learning set, on 3D-QSAR model on SERT (R2=0.96; Q2=0.89).

ADMET results
Results of in silico ADMET studies show that all te natural compounds present a high Caco2 permeability (all the compounds have a value higher than 0.90 log Papp in 10-6 cm/s, that is considered high).
The natural compounds have a medium, respectively medium-high CNS permeability, in case of resveratrol (-2.09) on the other hand amitriptyline and sertraline are higher -1.41, respectively -1,15. a value higher than -2 is considered that penetrate the nervous system.
Natural compounds and sertraline are not CYP2D6 substrate, but amitriptyline is. Melatonin, resveratrol, amitriptyline and sertraline inhibit CYP1A2 but linalyl acetate shows no inhibition. All compounds except amitriptyline are not renal OCT2 substrate.
The maximum tolerated dose (human) is similar between all the compounds, linalyl acetate has the highest maximum tolerated dose (0.54 log mg/kg/ day) and melatonin has the lowest one (0.38 log mg/kg/day).
No natural compound, present hERG I, or II inhibition, unlike sertraline and amitriptyline which both present hERG II inhibition.
The ADMET results for resveratrol, linalyl acetate melatonin and clinically used drugs in treatment of depression sertraline and amytriptyline are presented in Appendix A.

CONCLUSIONS
Based on the high power of prediction of our 3D-QSAR model (R2: 0.96; Q2: 0.89) our study conclusions reveal that melatonin, resveratrol and linalyl acetate, have similar biological activity on SERT with commonly used medication in treatment of depression (amitriptyline and sertraline).
ADMET in silico studies indicate that melatonin, linalyl acetate and resveratrol are well tolerated by human body in terms of absorption. Regarding toxicity melatonin resveratrol and linalyl acetate shows no hERG II inhibition, on the other hand amitriptyline and sertraline show inhibitory effect on hERG II.
However, the disadvantage of the natural compounds compared to the clinically used medication is represented by the lower CNS permeability.
Melatonin and resveratrol can be utilized in pregnancy, without the negative effects on foetus or new-borns.
The general conclusion of our study is that melatonin, resveratrol and linalyl acetate can be a reliable substitute to classical antidepressants in case of pregnant women that are suffering from depression.
The natural compounds are well tolerated by the human body, present fewer side effects, and are safe for the foetus and new-born.