Can We Learn To Distinguish between “Drug-like” and “Nondrug-like” Molecules?

Abstract

Lipinski's "Rule of Five" was introduced for predicting oral bioavailability to describe drug-like molecules. For the purpose of this research the rules were used to separate potential inhibitors of HIV-1 integrase (1BIS.pdb) into two groups: drug-like and nondrug-like. If one of Lipinski's "Rule of Five" was not followed the potential inhibitor was classified as nondrug-like. Thirty molecules were identified from the literature, twenty-four drug-like and six non-drug-like, that were docked into the active site of 1BIS.pdb (considered the non-mutated protein) and two mutant models, Y143R and N155H. These are two of the mutations that have led to increased resistance to HIV-1 integrase drugs such as raltegravir and elvitegravir. The computational software, ICM-Pro (Molsoft L.L.C.), was used to determine the estimated binding energy (EBE) of the drug/protein complex. It was found that the nondrug-like molecules generally had a more negative EBE, that is, tighter binding with 1BIS. pdb, though there were several exceptions in the drug-like group. With the protein mutant model Y143R, the majority of drug-like (58%) and nondrug-like molecules (67%) had tighter binding. However, for the mutant model N155H, there was the same percent (46%) of drug-like molecules with tighter binding with the mutant model as with 1BIS.pdb. The drug-like molecules were used when there was a ≥1 kcal/mole difference between 1BIS.pdb and either of the two mutant models to suggest a pharmacophore with structural characteristics for an HIV-1 integrase inhibitor.

An interactive database to explore herbicide physicochemical properties

“…This is in line with the proposed value of maximum polar surface area allowed for 35 bioavailability of ≤140 Å 219 . For the rotatable bond count it was found that in general the number of rotatable bonds increased with molar mass, consistent with what has been observed in libraries of pharmaceutical-type molecules 20 .…”

An interactive database to explore herbicide physicochemical properties

“…This is in line with the proposed value of maximum polar surface area allowed for 35 bioavailability of ≤140 Å 219 . For the rotatable bond count it was found that in general the number of rotatable bonds increased with molar mass, consistent with what has been observed in libraries of pharmaceutical-type molecules 20 .…”

“Drug-Likeness” <em>versus</em> “Natural Product-Likeness”

“…This model could correctly predict 77% of the WDI drugs and 83% of the ACD molecules as drugs and non-drugs, respectively. Similar results have been obtained using neural network [71,72] and decision trees [73], using the same databases of drugs and non-drugs and the same set of descriptors. The performance of the decision tree model was comparable to that of the neural networks, correctly predicting ~83% of a validation set not included in the initial model.…”

Similarity-driven flexible ligand docking

“…Comparison of the results obtained for the three different databases by each of the methods reveals that the percentage of active molecules retrieved within the 10% best‐ranked molecules tends to decrease from the ACD, to the DIV, and to the WDI database. This trend correlates well with the recognized degree of drug‐likeness of the molecules in each database 57, 58. On the one hand, the ACD database is assumed to contain very dissimilar molecules with nondrug‐like characteristics.…”

An interactive database to explore herbicide physicochemical properties

“…This is in line with the proposed value of maximum polar surface area allowed for 35 bioavailability of ≤140 Å 219 . For the rotatable bond count it was found that in general the number of rotatable bonds increased with molar mass, consistent with what has been observed in libraries of pharmaceutical-type molecules 20 .…”

“Drug-Likeness” <em>versus</em> “Natural Product-Likeness”

“…This model could correctly predict 77% of the WDI drugs and 83% of the ACD molecules as drugs and non-drugs, respectively. Similar results have been obtained using neural network [71,72] and decision trees [73], using the same databases of drugs and non-drugs and the same set of descriptors. The performance of the decision tree model was comparable to that of the neural networks, correctly predicting ~83% of a validation set not included in the initial model.…”

Similarity-driven flexible ligand docking

“…Comparison of the results obtained for the three different databases by each of the methods reveals that the percentage of active molecules retrieved within the 10% best‐ranked molecules tends to decrease from the ACD, to the DIV, and to the WDI database. This trend correlates well with the recognized degree of drug‐likeness of the molecules in each database 57, 58. On the one hand, the ACD database is assumed to contain very dissimilar molecules with nondrug‐like characteristics.…”

An interactive database to explore herbicide physicochemical properties

“…This is in line with the proposed value of maximum polar surface area allowed for 35 bioavailability of ≤140 Å 219 . For the rotatable bond count it was found that in general the number of rotatable bonds increased with molar mass, consistent with what has been observed in libraries of pharmaceutical-type molecules 20 .…”

“Drug-Likeness” <em>versus</em> “Natural Product-Likeness”

“…This model could correctly predict 77% of the WDI drugs and 83% of the ACD molecules as drugs and non-drugs, respectively. Similar results have been obtained using neural network [71,72] and decision trees [73], using the same databases of drugs and non-drugs and the same set of descriptors. The performance of the decision tree model was comparable to that of the neural networks, correctly predicting ~83% of a validation set not included in the initial model.…”

Similarity-driven flexible ligand docking

“…Comparison of the results obtained for the three different databases by each of the methods reveals that the percentage of active molecules retrieved within the 10% best‐ranked molecules tends to decrease from the ACD, to the DIV, and to the WDI database. This trend correlates well with the recognized degree of drug‐likeness of the molecules in each database 57, 58. On the one hand, the ACD database is assumed to contain very dissimilar molecules with nondrug‐like characteristics.…”