 It's okay. Do you hear me? Yes, yes, yes, we can hear you. I was just going to announce your title of this talk. So, Professor Pajevi is going to present us a report on ligand and structure-based studies of natural flavonoids from the plant Silibum Marianum. Thank you. Exactly. Okay. Okay. So, first of all, I'd like to know, do you hear me? You hear me? And do you see my presentation? We see it. It's not on full screen yet. It's just a moment. Yes. Okay. So, good morning or good afternoon to everyone depending on the time zone. First of all, I would like to thank the organizers and especially Professor Anelio Ivanova for giving us the possibility to show you our results as a success story in virtual drug development. As the title of the presentation shows, the study will discuss not only structure-based methods, which are the topic of the school, but also some ligand-based methods. And we hope that it will be interesting for the participants. The authors of the presentations are all members of the QSAR and Molecular Modeling Department in the Institute of Biophysics and Biomedical Engineering of the Bulgarian Academy of Sciences. And let us start. The presentation is structured as a classical study, starting from the rational, formulating some questions, then a description of the data and the methods, then the results will be present by giving answers to the questions set above. This will be done by three case studies. And finally, some conclusions will be drawn to show the outcome of our results. What is the rational of our study? The title of the slide shows this is the importance of natural product for drug development, particularly the biative natural products or a source for medical treatment and preparations from millennium, but they are very much used also in the modern drug discovery. Here you see the pie taken from a very recent publication showing how the new approval in the 40 years, past years, worldwide, that means not only foot and drug administration agency data, are distributed according to their type. And one can easily calculate that about 50% newly drug approval in some way are related to natural products. They are either natural products or their derivatives, or they mimic the function of natural product or contain a natural product pharmacophores. This comes to show how important are natural products for the drug discovery and development. The next thing that motivated us is the fact that the medicinal plants in Bulgaria, that Bulgaria is rich in medicinal plants. We have about 750 pitches, 200 of them are currently in use, and 30 to 40 medicinal and aromatic plants are cultivated. Six of them are most important among those cultivated, and among them is the plant we are interested in. This is Silibum mariano, or known in the folk medicine as the mule pistil. So what is famous this compound with? The seeds and the roots of the compound are used to get the so-called Silimerine extract, which standard application is for treatment of liver diseases. So it roses chronic hepatitis, diseases associated with increased alcohol consumption and exposure to environmental toxins. The new and emergent applications of this extract reveal a number of new effects to mention several, anti-cancer, cardio and neuroprotective treatment of different diseases of lung, pancreas, prostate and kidney. All these effects particularly put several questions when using this Silimerine extract. And what were the questions we put to ourselves? Using in silico methods, can we understand more about admit properties of the Silimerine compounds? We focused on the prediction of gastrointestinal absorption, and this was our case study one. And the second question was using in silico methods, can we understand more about the mechanism faction of the Silimerine compounds? And in this case we focused on the interaction with and prediction of potential protein targets in the human organism of the Silimerine compounds. And this were our case studies two and three. The abbreviation admit, you know from the previous lectures, this stands for absorption, distribution, metabolism, excretion and toxicity associated with the pharmacokinetic stage of the drug action. As any in silico study, this one starts with defining data. Silimerine is a mixture of compounds, but in silico we need a defined structure. And here you see the main active ingredients in Silimerine. Those are seven compounds among them, Silibene A and Silibene B, which are deaesteroisomers. Also, they are dehydroderivatives, dehydro-silibene A and B. These are mostly representatives of the class of flavonoidinans, but we have also some flavonoids like taxifoline. The main methods, they are in silico or computer-aided drug design as stated in the title, represented as a ligand and structure-based methods. Each group has its advantages and disadvantages. And if one can combine both methods when possible, one can really do more effective research. Among the ligand-based methods, we use Thucer modeling, also virtual ligand-based screening estimating similarity between compounds, between ligands. And among the the structure-based methods, we use docking. All these methods will be given, for these methods will be given more details in the further presentation. Let us start with the case study one. We aimed at the prediction of gastrointestinal absorption and focused in particularly on defining or calculating the membrane permeability of the compounds having in mind that the membrane permeability correlates very well with the gastrointestinal absorption. The technique or the method we used is the QSAR model. Most of you know that in general it correlates any activity or property data of the ligands with structural descriptors or molecular descriptors of the compounds. Different kinds of multivariate statistical methods can be used to draw these dependence, but in this case we used the multiple linear regression of two parameters. Here you see a schematic presentation of the idea. If we have two independent variables x1 and x2 using the least square method, we are trying to build a plane that goes, that passes simultaneously most close to all points. These are the projection of the points by x1 and x2 in the space. In our case the dependent variable, the y value, was the effective permeability coefficient measured by PAMPA methodology. This has also been mentioned in the previous lecture. These are parallel artificial membrane permeability assays. Then log g, the distribution coefficient at pH 7 and the topological polar surface area and molecular weight have been used as structural descriptors. The distribution coefficient we calculated by in two ways, by chemaxon marveline and also using the fiscam suit of Perceptor. The topological polar surface area and molecular weight were calculated by the CDK9 nodes. The equation itself in its complete form with all statistical results is shown on the slide. It is based on data of about 250 compounds. You see that we have relatively good statistical parameters, I would say even very good. We calculated not only the fitting coefficient presented here by the adjusted R-square, but also the predictive coefficients by Rikwan-Auth method and by splitting the data set into the training and test set correspondingly shown on the plot in green and red. The plot presents the dependence between the experimental and calculated by the model log P values. You can see that there is a very good correlation between both. We further tested the capability of model to predict GIA assessed on a set of 780 compounds with known gastrointestinal absorption. Where are our 7 compounds from the Cilimarine in this model? It is shown here again the same plot and in black dots you see our compounds. So they have been experimentally tested also to predict their membrane permeability. You see that the compounds fall into the applicability domain of the model and 5 out of 7 go to the half of the model which shows high permeability of this membrane permeability of these compounds. This model gave us a hope that we can also predict the membrane permeability of non-tested flavonulignans and we did this for 31 compounds and for 27 of them we found good permeability and high permeability in the gastrointestinal tract but this hasn't been yet experimentally proven. The next study relates to the interaction of our compounds with the nuclear estrogen receptor alpha. In fact we were interested in toxicity of these compounds due to their broader use and performing silico prediction using the Derek Nexus expert system and the result was that the Cilibin A and Cilibin B are able to modulate the function of the estrogen receptor alpha in mammals without of course giving any differences in the prediction of their effects and thus we decided to perform talking in the binding site of the estrogen receptor alpha. Here you see the ligand binding domain with bound endogenous ligand estradiol and the activation helix 12 which is very important for the function of this receptor in the agonist and antagonist conformation. We used two softwares moe and gold. The docking was semi-flexible solvent was considered semi-flexible docking in this case means rigid receptor and flexible ligand but I have to say that we also allowed the side chains of the amino acids facing the pocket to be also flexible and here I would like to draw your attention on the fact that two positions of the activation helix were taken into account why. I think that in all structure-based studies it is very important to pay attention to what crystal structure, the structure of protein or enzyme we are using because they could be different depending on the functional state they are. This is especially important for the nuclear receptors and also for the trans proteins. The binding site was defined by the ligand. Here I give the data about the PDP complexes used and the scoring functions were London VG in moe and gold score in gold. Both are empirical scoring functions that consider not only contribution to the enthalpy but also to the entropy of the interaction energies. I think that in the previous presentation in the first one was given the formula for the London VG I will not go into the detail more about. We did observe some differences in the interactions of our syllabine A and B in the binding site. This is illustrated here and the differences in the poses suggest some stereo specific interactions. Interestingly, in most cases we recorded hydrogen bond interactions of syllabine B and no specific interactions for syllabine A in most of the poses. Thus we were looking for the experimental confirmation of our results and we were very happy to find in this study that stated about syllabine B and not syllabine A to be responsible for the partial mediated activity in relation to the hydrogen receptor alpha of the polymer ring. In this way we could explain at a molecular level the experimentally observed differences in the activity of both syllabines on this nuclear receptor. The case study three is related to the discovery or to our trial to discover new protein targets of the syllabine compounds. The research strategy is based on the fundamental principle in drug action complementarity between the drug and its binding or receptor binding site in two dimensions shape similarity and shape complementarity and complementarity of the surface properties or surface potentials. The extrapolation of this idea of complementarity between the drug and its binding site could be given to the ligands only meaning that we can look for ligands that are similar in shape and also similar in their surface properties and we use the rock software under the open IPA academic license program to perform such similarity assessment. You see here on the left the shape correspondence and the correspondence in the surface properties is done by the so-called color similarity meaning identification of functional groups, atoms and substructures capable of performing similar intermolecular interactions like scepter, donor, hydrophobic, aromatic, ionic and so on. The similarity has been estimated using the TANIMOTO coefficient. The most simplified form of the TANIMOTO coefficient is shown here. The number of features of one molecule, of the other molecule, the common features between them, but when we try to estimate more complex properties like shape and functional similarities, the form is much more complicated. We have non-binary data. The features are presented at bits sets in the fingerprints and we calculated the TANIMOTO combo coefficient combining shape and color similarity together. Therefore, the coefficient ranges from zero meaning no similarity to two meaning full similarity. The slide shows the workflow of the search. We used all approved drugs from the drug bank to that time they were about 200, 2300 meaning we know everything about these drugs. We know their structures. We know their effects. We don't know their targets. What is very important and we found nine drugs for Cilibin and their dehydral derivatives to have TANIMOTO scores above 0.9 and two of them to be similar to drugs with anti-tumor activity. Then the question was can Cilibins and their derivative interact with the same targets as the similar anti-tumor drugs do? Here we summarize the values of the TANIMOTO combo indices and show what are those two drugs. One of them is Vemurofenyep and it is used for treatment of metastatic melanoma and its target protein is the protein kinase B-RAF and this modigip which is mostly used for treatment of basal cell carcinoma and the target protein is smooth and receptor. But before to proceed further with studying the interactions of our compounds with those two target proteins, we perform additional check using more software by flexible alignment of compounds on the drugs by active confirmations as extracted from their 3D complexes and mapped properties on their molecular, in this case, conical surfaces to see how they correspond to each other and this is seen here. On the left you see the shape alignment between dehydro-Cilibins in green and Vemurofenyep and Bismutigip in brown and on the right the surface properties mapped for Bismutigip and both Cilibins here we have electrostatic and hydrophobic hydrophilic properties. In general, we have better correspondence in shape for Bismutigip and relatively good correspondence in the surface properties, I would say good correspondence even in the surface properties between Bismutigip and Cilibins. We further, as I mentioned, perform docking in the active site of the beerhead kinase you see here the PDB complex we used and the table reports on the docking scores for the best score posts and the first three, the average of the first three of our four compounds and also Vemurofenyep because we redocked our compound into the binding site, our drug into the binding site to get an idea about the docking scores. You see that Cilibins and dehydro-Cilibins have lower docking scores compared to Vemurofenyep but still very good. Here you can see their binding in the active sites of Cilibin A and B and dehydro-Cilibin A and B. I have to say that we especially followed how the interactions compare to the interactions of Vemurofenyep and we found that our Cilibins and dehydro-Cilibins could do interact with the residues. Vemurofenyep interacts with it and in addition we recorded different interactions, especially evident for dehydro-Cilibins in this receptor but also in the other one. Based on that, we concluded that certainly these compounds can interact with B-RAF kinase but it was not enough and we decided to look for the individual validation of our results and we did perform experiments for inhibiting B-RAF kinase in its mutation which is in 80-90% of advanced melanoma patients present, valine to butamic acid in 600 positions and we found in accordance in correspondence with our N-Cilico results that dehydro-Cilibin has the highest inhibitory effect compared to the other compounds. The IC50 of dehydro-Cilibin was about 25 micromole. Further, we did docking similarly in the receptor of Smuton homolog. This is the G-protein coupled receptors, as I said, involved in the head-to-heart signaling pathway, very essential for the cancer diseases development and again we found good scores for our compounds compared to this modigip. We performed a strong validation of our N-Cilico results by two assays and we found very interestingly that dehydro-Cilibin A was active while dehydro-Cilibin B was inactive, both C-Libins A and B showed MUT activity and additionally we performed displacement assay for the best compound and we did find that these compounds bind to the Smuton homolog receptor. The main conclusions drawn from our case studies is that in silico methods are really useful. We could predict the membrane permeability in the gastrointestinal tract. We explained at the molecular level the specificity and effects when interacting with the estrogen receptor alpha and we proposed at the level of in vitro experiments proof that indeed these compounds could interact with two noble targets in the human organism. So we can conclude that in silico methods are very valuable tools when looking for a drug. Here are the publications related to our studies. They all have already been published. We have a lot of support from different programs also from international projects. They are also mentioned here and in the end we formulated three messages to you to take home because we do think that they are important and these are also based on our experiment experience. The first is do not forget essentially all models are wrong but some are useful. This has been said by George Box. So our advice is try to validate experimentally the in silico models and results. Second follow the principle of parsimony or Occam's razor meaning all things being approximately equal one should accept the simplest model and the third advice is no matter what computational method you are using don't get lost in calculations and keep your eye on the goal and in the end I'd like to thank for your attention. Thank you. I see that there's one question in the chat are those 248 old Slavo Nuits are there log p values and toward the log p values taken from literature or measure can you can you please I would like to read the question okay show it yeah well I cannot show it on your screen but you can simply stop sharing my screen so we'll read it again okay are those 248 old Slavo Nuits no no no not okay you mean you mean the model for membrane permeability no no they are different drugs no okay and were the log p values taken from the literature or measured all at once for the model we had measured membrane permeability coefficients otherwise it was not possible to build this plot of the predicted versus observed but for the silimony compounds we did indeed we did experimental measurements using pump assay and this has been shown also by their positions on the plot okay this was a kind of validation of our results okay how did you select those specific descriptors among the vast number of existing descriptors the common sense first of all uh the dependency yes the membrane permeability is a very complex property it depends on many manufacturers most of them are related to the steric and to the lipophilic or hydrophilic properties of the compounds and of course one can do also a lot of calculations using different descriptors performing principal component analysis dividing deriving different kind of transformation of the initial set of variables but we did we did get very good results with these simple models and with these simple descriptors that fully satisfied us and therefore we decided to rely on them okay thank you I thank you for the very interesting talk and now I suggest that we continue with the next presenter yes I would like only to mention that I have to leave now the session and if there will be any questions to us uh I promise to answer later on if you don't mind okay okay okay thank you thank you again