Відео

Thank you for your comment.

Dear, The datasets for QSAR modeling and other tasks are usually obtained either from the literature (i.e. a research article that reports a set of structures and their activity values) or from a public compound database such as the PubChem and ChEMBL databases. Those databases contain compounds and their reported biological activities. You can also contact us (crescent_silico@outlook.com), if you need ready-to use datasets for QSAR modeling. Regards

Unfortunately, we currently do not have a version for Mac. However, If you need Computer-Aided Drug Design services/coding you can contact us at: crescent_silico@outlook.com .

You can download the software from the official download page: crescent-silico.com/chemmaster

This is just to make sure no structure is present twice or more in the dataset. For example, sometimes when you download structures from a database, the same structure might be present multiple times, which is bad for model development.

Thank you for your comment & support.

Dear, Thank you for your comment. Ligands preparation is required before QSAR modeling as well, however, some of the steps used here are not always necessary. For example, generating 3D-conformers (i.e. 3D-coordinates) is required for 3D-QSAR only. If you are doing QSAR without 3D-descriptors, generating 3D-conformers is not required. Note that you can perform preparation in the ChemMaster software as well (Structure Menu -> Prepare Structures). Regards

Dear, Thank you for your comment. The HQSAR method doesn’t give an equation as the number of X-Variables (i.e. the descriptors which are the molecular fingerprints) is very large. Thus, in HQSAR research articles, the equation is not reported. As for R2, generally you should focus on the test set value, as it reflects the model ability to predict activity on new structures. The CV is for Cross Validation, which is an additional (optional) validation procedure. The R2 of CV is similar to Q2. We have a planned tutorial for the HQSAR module, so stay tuned if you are interested. Regards

do you have this version?

@alien2883 If you need this version or other products, please contact us & we'll provide you discount if you need.

@@CrescentSilico actually I am a student and I can't afford it. Thanks for the help

@alien2883 I can provide you a significant discount on the Dockamon 1.2 - Pro version, that should be absolutely affordable, if you are interested please contact us at crescent_silico@outlook.com

Dear ATEEQALALI, Thank you for your comment. The activity values such as EC50 or IC50 for the structures are usually obtained from public compound databases such as PubChem or ChEMBL. In such databases, you can find structures and their reported activities. You can also get datasets (structures & their activity) from the literature (e.g. research articles), which reports the structures and their activities.

Thank you for reply. I am appreciating for you. I would be grateful if could let me know what is the point to add an enzyme as a target? Can i build a model without adding the enzyme to the dataset?

Dear, The enzyme was not added to the dataset. The dataset contains only the chemical structures (small molecules). The enzyme is the drug target for all the structures (i.e. the structures act as an inhibitors for the enzyme), hence the target column only clarifies what is the drug drug target for those compounds. The activity values/class of each compound represent its ability to inhibit that enzyme.

@@CrescentSilico thank you for your comments. It will be very useful if you could make another video to interpret the model plot and explain it. Thank you dr

You're welcome. For binary-QSAR models, plots are not commonly used and reported with the model as in regression (numerical)-QSAR.

Dear Mohammed, You can use any activity type (including mIC90) to build a QSAR model (you may consider scaling the activity as well as shown in the tutorial). The built model will predict the same activity type that was used for training (thus if you use mIC90, the model will predict mIC90 values for new structures.

Dear, For the current version, no installation process is required. After you download the ChemMaster.Rar file, right click and select Extract Here. Then open the ChemMaster folder and run the ChemMaster.exe file. If you are still unable to run the software, please contact us at crescent_silico@outlook.com and we'll send you detailed steps if required.

@@CrescentSilico thank you for the timely assistance.

Dear, This version is commercial, however, we can give you considerable discount, if you are interested, contact us via email: crescent_silico@outlook.com

Dear, Thank you for your comment. Currently, ROC curves are not provided in ChemMaster, however, we can generate it for you using other tools we have in CrescentSilico, if you can provide us more context regarding what data you want to generate ROC curve for? Generally, regression QSAR models (as in this tutorial) are not validated by ROC curves. You can contact us at: crescent_silico@outlook.com

Dear, The selection of the ranges for activity classes can be indeed a difficult task. There are no specific calculations that can be made to ensure proper results. However, there are two main points to consider. First, try to select ranges that result in balanced classes as much as possible (i.e. classes containing approximately the same number of compounds). Second, the ranges should reasonably represent each class (e.g. highly active, low-active), this depends on the range and distribution of the activity in the dataset. If you are interested, we offer services for QSAR modeling: crescent-silico.com/cadd-services/

Dear, The causes for a low R2 can result from various factors. For example, the characteristics of the dataset you are using (activity range, activity distribution, no. of points, etc.), or the type of descriptors used. If you like, you can contact us via the CADD services page, we offer extremely low-cost services, with experts in QSAR who can work on your dataset and/or provide you suitable datasets for QSAR (crescentsilico.wordpress.com/cadd-services/).

Dear, If you mean your IC50 column name is not in the "Add Column" list choices in the Add Calculated Column Dialogue, that should be because the type of your column is not "Numerical" (to set a column datatype as Numerical, right click on the column name -> Set Column Datatype -> Numerical). Note that the column should contain only numerical values to be set as numerical.

Dear, If you want to save the image of the figure, you can do so by right clicking on the tab name of the plot and selecting "Save Image". To change the background color of the plot, click on the plot options icon (the three vertical blue lines) -> General -> Background Color. If this doesn't help, please elaborate more on the issue. Thank you.

Thanks for the swift reply @@CrescentSilico

Dear, The purpose of a random seed is to make sure the random division is the same (repeatable), the value itself is meaningless. In other words, when you enter the same random seed you get the same compounds in the training/test sets every time. It is just to control the random process to get reproducible results. You can enter any value, but each value will give you different dataset division (i.e. training/test set ids), and each value always results in the same dataset division. I hope this is clear.

Dear, The minimum and maximum are the range of the class. In the tutorial, they are selected according to what you want to consider as "Highly Active" / "Low-Moderatly Active". Here, we considered 100 nM or less IC50 as "Highly Active" and the rest as "Low-Moderatly Active". You can set a different threshold as well, for example you can consider 1 μM or less IC50 as highly active, in such a case the "Highly Active" class will refer to compounds with 1 μM or less IC50.

Thank you sir for your answer . but I have another question what does we mean by testing and training and what is the purpose of dividing our compound to this groups

Dear, Training refers to building the model (i.e. the learning process), the model can be then used to predict new data with unknown activity. However, before using the model, we need to make sure that the model predictions are accurate and reliable, we do this by testing the model on a different set which is the test set. Then we compare the predictions of the model with the real activity to assess the accuracy and validity of the model. You can also consider getting our recent course on QSAR modeling, which explains each concept in details: udemy.com/course/qsar-modeling-principles-and-practice/?referralCode=DA624B915FB6C1419B85

Hi Omar, newbie in QSAR here. How do I run QSAR analysis of 4 synthesised molecules with determined Kd values?Thanks

Dear, Generally, 4 molecules are considered to be a very small dataset for QSAR modeling and impractical. You can add other molecules to the dataset (e.g. from the literature/databases), as the molecules don't have to be synthesized by you for performing QSAR. You can also consider performing other CADD methods on your molecules such as pharmacophore modeling, as the number of molecules is acceptable for it.

Yes, you can import descriptors from other sources. To do so, from the file menu select "Import Columns" then select the file. The current formats are CSV (comma separated) and Text Files (tab separated), and the first row is expected to be the header name. If this doesn't work, you can contact us to try to provide better solution.

Thank you for your suggestion, that is something we have been considering. We will attempt to do that in the near future.

Dear, if you mean the SDF file used in the video, you can find it in the description. However, if you have your own dataset of compounds, you can simply draw the structures via a software or webserver that permits drawing the structures and export them as an SDF file. If you can give us some context about what you are using and what you want to achieve, we can provide better help. Thank you.

I mean in general How to get collectively structure along with Ec 50 which software used for that, I know how to get sdf file for structure

Dear, if you want to get a dataset for a QSAR study, you can get a series of structures from a literature article that reports the structures with their activity. Or a simpler way is to download a dataset from a public database such as the ChEMBL database. If you want a dataset suitable for a QSAR study, you can contact us via the CADD services page, we can assist you in any step in QSAR modeling including dataset selection (crescentsilico.wordpress.com/cadd-services/).

The SDF file contains the structures and the activity of your dataset, typically, you may draw the structures via a webserver or a software that permits drawing the structures and export them as an SDF file. If you can give us some context about what you are using and what you want to achieve, we can provide better help. Thank you.