The manuscript is interesting one. The title and abstract provided the aims and ideas of the work clearly. The authors had provided relevant and up to date references. I believe the work can be considered for publications after responding to the below-mentioned queries.  Regarding the dataset, the dataset is prepared employing three different bioassay ID from PubChem. There are two queries: 1. Have authors checked the similarity of biological assay protocol for three different dataset? Without a similar protocol, one can’t merge three different datasets for QSAR modeling purpose maintaining OECD principle 1. 2. I checked PubChem and found 70, 14 and 14 active compounds under AID 319991, AID 268043 and AID 162506, respectively. So, why dataset consists of 72 molecules instead of 98?

 Regarding the UNPD dataset, I have checked this reference [19. Gu J, Gui Y, Chen L, et al (2013) Use of Natural Products as Chemical Library for Drug Discovery and Network Pharmacology. PLoS One 8:1–10. doi:10.1371/journal.pone.0062839] as well as I have checked this link (http://pkuxxj.pku.edu.cn/UNPD) but can’t access the dataset. So, please provide the correct link for the dataset or provide the whole dataset used in this study in Supplementary materials as single .sdf file. It will be helpful for readers.

 I haven’t found any randomization study results for pharmacophore study. Please provide.

 In Figure 10, amino acid residues are not clear at all. Even Schrödinger generated docking figures that are unreadable. Please make it clearer. You can provide clear figure for the best lead compound and the remaining ones can be transferred to the supplementary file.  Similarly, in figure 14, the amino acid residues are unclear.  Usage of multiple structure-based algorithms and their results must be properly justified with a specific paragraph under the discussion section.
 The conclusion is quite general. The author must conclude with 1-2 lead inhibitors or candidates for COX-2.

Recommending major revision.