Hi guys
I have a question related to descriptor selection for QSAR modeling. We are using a software called PADEL for calculation of 2D and 3D descriptors. We can import our molecules in different formats such as sdf, mol2, mdl and smi,but we observe different values for the same descriptors in different formats. I am really thankful If anyone can help us.
Looking forward to hearing from you.
Samane
Dear Samaneh,
This problem is widely known. For the reasons, see the following link:
https://books.google.it/books?id=LZt3PXHBzboC&pg=PA61&lpg=PA61&dq=different+values+for+the+same+descriptors+(descriptor+selection+in+QSAR+modeling)+in+different+formats&source=bl&ots=6y8GcFzA6X&sig=aAFzQppdSz45zZMKXfp6FftreMw&hl=en&sa=X&redir_esc=y#v=onepage&q=different%20values%20for%20the%20same%20descriptors%20(descriptor%20selection%20in%20QSAR%20modeling)%20in%20different%20formats&f=false
Hoping this will be helpful,
Rafik
Dear Rafik
Thank you for your reply. Is there anyway to minimize this source of error? Which format is better to use?
Dear Samaneh,
in addition to what already reported by Prof. Rafik Karaman I would like to say that the same molecule can be represented using any kind of molecule file format. If the molecule (independently from the file format) has the same atoms and connections it should generate exactly the same descriptor values.
I am not familiar with Padel, you should check if Padel performs molecule standardization or not. If it does you should obtain the same descriptor values, otherwise descriptor values could be different. e.g a benzene represented using kekulè form (3 single and 3 double bonds) or aromatic form (6 aromatic bonds).
If you provide a pair of molecules represented using two different file formats and giving different values of molecular descriptors I can check it.
With respect to the molecule file format, the most used file formats are SMILES if you don't need 3D information and MDL/SDF if you have 3D information.
Andrea
Dear Andrea,
Many thanks for your reply. You are right. As you know the best software for descriptor generation is Dragon. Since it is not free, I was forced to use another software. I think there was a problem related to "Padel". I am using another software called "chemopy" ; Although the number of descriptors are less than Padel or Dragon, the results do not change with different formats. We could get satisfactory results for QSAR model.
Thank you again
samane
Dear Samaneh,
take into account that a free version of Dragon is available trough VCCLAB website (http://www.vcclab.org/lab/edragon/). It is based on an old version of Dragon (dragonX 1.4) and it has some limitations on number of molecules per batch and number of atoms per molecule but it could be a good starting point.
Andrea
Dear Andrea,
Thanks a lot for this Website Address. Although my QSAR model with Chemopy is acceptable, I will check this website too. I think the number of descriptors in this version of Dragon and Chemopy are similar.
The reason why could this happen has been answered by Andrea Mauri. Thank you for choose our "Chemopy" to calculate molecular descriptors, and we have developted a free integrated web-based platform for molecular descriptor and fingerprint computation: ChemDes(http://www.scbdd.com/chemdes) and a web-accessible platform for generating various molecular representations for chemicals, proteins, DNAs/RNAs and their interactions: BioTriangle(http://biotriangle.scbdd.com). These tools are free and need only "upload" and "download".
- Badges
- Science topic
- Similar topics
- Biological Science
- Bioscience
- Biotechnology