The second similarity measure is the basic local alignment search tool (BLAST) string matching method which is used to evaluate the text-based molecular descriptors (SMILE, LWDOSM, and GBMD).
In this section, we conduct experiments that show the usefulness of our proposed descriptor GBMD when used for similarity-based virtual screening.
The experiments are conducted to identify the possibility of using the GBMD descriptor in similarity-based virtual screening and then identifying the retrieval effectiveness of using such a descriptor.
Visual inspection of the recall values and the number of bold cells in Tables 4-9 enables comparisons to be made between the effectiveness of the GBMD descriptor and the various other descriptors.
Here, it is shown that the GBMD descriptor has the best overall performance at the 1% and 5% cutoff.
The results in Tables 6 and 7 show that the performance of GBMD is inferior to the best performance descriptors (CDKFP and EPFP4).
To determine to what degree GBMD descriptor obeys this principle, a GBMD descriptor was used to search for the most diverse set of active classes (DS3 data set).
Visual inspection of the results in Table 11 shows clearly that the GBMD descriptor can provide a level of performance that is generally superior to the other descriptors.
Retrieval results of top 1% and 5% of GBMD for data set DS1-DS3 compared with the LWDOSM and Lingo-DOSM are shown in Tables 12-14.
From the above results, it should be noted here that the main purpose of using several types of descriptor in the experiments was not a performance comparison only but also to show that our new descriptor GBMD is capable of representing and characterising the molecule structure and to show the possibility and feasibility of using GBMD for similarity-based virtual screening.