I have built the 2D QSAR with BIOVIA DS and got r2 =0.9516 and q2 = 0.907, but i am still confused that is my graph is correct or not because the test set doesn't spread in under the trendline.
Looking at the chart it seems that the model is not so good. It fits well your training data but it overestimates the PIC50 for the compounds in the test set.
You could try splitting the training and the test set in a different way.
Additionally, I suggest you to check our QSAR tools at:
https://www.alvascience.com/
You can also check our last paper describing the whole QSAR workflow:
Article Alvascience: A New Software Suite for the QSAR Workflow Appl...
Thank you so much Andrea for your response. I'll try to check my QSAR in your tools.
so, although the q2 > 0.5 (in my QSAR is 0.857) they became not a good value because they are overestimated in the test set data
?
The Q2 value is good. My concerns are related to the fact that (according to the chart) the model works well on the training set but not so well on the test set.
I suppose that Q2 and R2 you reported have been calculated on the training set. What's the value for R2 calculated only on the test set?
- Badges
- Science topic
- Similar topics
- Filing