Hi,
I have set of 200 molecules along with their biological activities from same laboratory. However all these 200 molecules are structurally diverse. Further i wanted to carry out some molecular modelling study to determine trend or pattern of physicochemical properties with that of biological activity.
Can any one suggest a methodology which i can use to find out trend/correlation between biological activity and physicochemical properties of these molecules.
Thank you.
From my point of view Machine Learning will be the best approach in this case. If you have already performed QSAR/QSPR then use the same molecules as training set and after training use them as test set against each other. Use WEKA (Waikato Environment for Knowledge Analysis). It is developed with great tools for such workflow generations.
I think for these 200 molecules you can use the calculated physiochemical properties as descriptors as independent variable and take biological activity (pIc50, LD50, etc.) as dependent variable by 70% training set and 30% test set to make QSAR Model and find the correlation between the Experimental biological activity vs. Observed biological activity.
Thank you Rajnikant and Vijay,
Well i have already identified a series of analogous compound and carried 3D QSAR for them. Here i found two series one contain 50 molecules and another 20 molecules. Interestingly i got similar contour map for all interaction field (Hydrophobic, steric, electrostatic, hydrophobic etc) for both series.
Apart from this i have developed one pharmacophore for all structurally diverse active molecules. Further i have designed some molecules form developed pharmacophore and contour map of 3D QSAR.
Now the question start here. Apart from these Pharmacophore and 3D QSAR contour map i wanted to find out cut off values for physicochemical properties as new filtering criteria for newly design molecules, as newly design molecules have very scattered physicochemical properties.
Data Description= In these 200 molecules 50 molecules are highly active, 100 molecules shows intermediate activity and 50 are inactive. I found positive charge is important for activity from pharmacophore and 3D QSAR. I have already check proper ionization state by using Chem Axon software.
Please guide me how i can generate physicochemical property filters from this data.
Well i am thinking of cluster analysis, will it help me??
Thanks,
Chintamani J.
- Badges
- Science topic
- Similar topics
- Computer Science and Engineering
- Bioinformatics
- Bioinformatic Tools