Drug-like indices

dProperties calculates 27 drug-like indices. These are dummy variables taking value equal to one when all the criteria of the consensus definition of a drug-like molecule are satisfied, 0 otherwise. These are filters used to extract good drug candidates from large collections of compounds.

A drug-like score is a real value ranging from 0 to 1, calculated as the fraction of criteria satisfied: a score of 1 indicates that a compound is a good candidate to be a drug, whereas a score of 0 indicates that a compound will likely not be a drug.

Lipinski Rule of 5 (Ro5), also known as Lipinski Alert Index, is a filter that identifies compounds with low probability of useful oral activity because of poor absorption or permeation [C.A.Lipinski, F.Lombardo, B.W.Dominy, P.J. Feeney, Advanced Drug Delivery Reviews 1997, 23, 3-25; C.A.Lipinski, F.Lombardo, B.W.Dominy, P.J. Feeney, Advanced Drug Delivery Reviews 2001, 46, 3-26].

In the discovery setting the Lipinski rule of 5 predicts that poor absorption or permeation is more likely when:

This computational alert is a filter that identifies compounds lying in a region of property space where the probability of useful oral activity is very low. A compound that fails the alert will likely be poorly bioavailable because of poor absorption or permeation [C.A.Lipinski, F.Lombardo, B.W.Dominy, P.J. Feeney, Advanced Drug Delivery Reviews 1997, 23, 3-25; C.A.Lipinski, F.Lombardo, B.W.Dominy, P.J. Feeney, Advanced Drug Delivery Reviews 2001, 46, 3-26].

This alert index is a dummy variable taking value 1 when two or more properties are out of range.

The 'H-bond donors' property is determined by adding up all of the hydrogens bonded to Os and Ns without negative charge. The 'H-bond acceptors' property is determined by adding up all of the Os and Ns in the molecule. The Moriguchi MlogP is calculated according to the rules implemented in dProperties, which partly differ from those proposed in the Lipinski's papers.

Drug-like scores (DLS) and Lead-like Scores (LLS) are defined as the ratio between the number of satisfied conditions over the total number of conditions.

The index DLS_01 is a drug-like score based on the Lipinski’s rules.

The index DLS_02 is a drug-like score based on rules similar to those implemented in the drug-like filter of Oprea et al. [T.I.Oprea et al., J. Mol. Graph. Model. 2000, 18, 512-524; T.I.Oprea, J. Comput. Aid. Mol. Des. 2000, 14, 251-264]:

The index DLS_03 is a drug-like score based on rules similar to those implemented in the drug-like filter of Walters et al. [W.P.Walters and M.A.Murcko, Adv. Drug Deliv. Rev. 2002, 54, 255-271]:

The index DLS_04 is a drug-like weighted score based on rules similar to those implemented in the drug-like filter of Chen et al. [G.Chen et al., J. Comb. Chem. 2005, 7, 398-406]:

C3p is the ratio of the number of C(sp3) atoms over the total number of non-halogen heavy atoms; h-p is the ratio of the number of hydrogen atoms over the total number of non-halogen heavy atoms;

Unsat-p is the ratio of molecular unsaturation, as defined by the Unsat index, over the number of atoms which do not have bonded hydrogens and halogens.

The Unsat index [S.Zheng et al., J. Chem. Inf. Model. 2005, 45, 856-862] is calculated as:

where NRG567 is the number of 5-, 6-, and 7-membered rings, nDB the number of double bonds, nTB the number of triple bonds, and nAB the number of aromatic bonds.

The index DLS_05 is a drug-like score based on the two rules proposed by [S.Zheng et al., J. Chem. Inf. Model. 2005, 45, 856-862]:

where Unsat-p is a measure of molecule unsaturation and is defined as for the index DLS_04, NO_C3 is an index related to the proportion of heteroatoms, defined as the ratio of the total number of oxygen and nitrogen atoms over the number of carbon atoms with sp3 hybridization.

The index DLS_06 is a drug-like score based on rules derived by the filter proposed in [G.M.Rishton, Drug Discov. Today 2003, 8, 86-96]:

The index DLS_07 is a drug-like score based on the two rules of the filter proposed in

[D.F.Veber et al., J. Med. Chem. 2002, 45, 2615-2623]: a) number of rotatable bonds (RBN) ≤ 10, and b) polar surface area (TPSA(tot)) ≤ 140 Å2 or the sum of  H-bond acceptors and H-bond donors ≤ 12.

Dragon consensus drug-like score (DLS_cons) accounts for the results provided by all the implemented drug-like scores (DLS_01 to DLS_07), it being calculated as their mean.

While the term ‘drug-like’ is used for compounds resembling existing drugs, the term ‘lead-like’ for compounds possessing the structural and physico-chemical profile of a quality lead. Lead-like scores are filters used to select those compounds qualified to be a lead in drug discovery. Compared to drugs, leads have, on average, smaller molecular complexity (smaller molecular weight, less rings and rotatable bonds), smaller polarizability, are less hydrophobic (their logP is 0.5 – 1.0 units less than that of drugs), and have lower drug-like scores [M.M.Hann et al., J. Chem. Inf. Comput. Sci. 2001, 41, 856-864]. Therefore, in general, physico-chemical property values used as a measure of lead-likeness should be smaller than those traditionally used for drug-likeness.

dProperties provides two lead-like scores derived from filters proposed in the literature.

The index LLS_01 is a lead-like score derived from the rules proposed by Congreve et al. [M.Congreve et al., Drug Discov. Today 2003, 8, 876-877]:

The index LLS_02 is a lead-like score derived from the rules proposed by Monge et al. [A.Monge et al., Mol. Div. 2006, 10, 389-403]:

where the number of H-bond donors is determined by adding any heteroatom with a minimum of one hydrogen and without negative charge, and the number of H-bond acceptors is determined by adding any nitrogen, oxygen, phosphor and sulfur, except for aromatic oxygen and sulfur, aromatic nitrogen connected to three other atoms, nitrogen with valence 5, and sulfur with valence 6 or 7.

Eight drug-like indices were proposed by Ghose-Viswanadhan-Wendoloski [A.K. Ghose, V.N. Viswanadhan and J.J. Wendolowski, J. Comb. Chem. 1999, 1, 55-68] in order to help to streamline the design of combinatorial chemistry libraries for drug design and develop guidelines for prioritizing large sets of compounds for biological testing. They are based on a consensus definition and have been derived from analysis of the distribution of some physicochemical properties (logP, molar refractivity, molecular weight, number of atoms) and chemical constitutions of known drug molecules available in the Comprehensive Medicinal Chemistry (CMC) database and seven drug classes defined by desease state.

Among the eight proposed indices there are a general drug-like index, named CMC in dProperties, which has been derived from the analysis of the whole CMC database, and seven specific drug-like indices derived from the analysis of the considered drug classes (see the table below).

LogP (AlogP) and molar refractivity (AMR) are calculated by using the atomic contribution method of Ghose, Crippen and Viswanadhan implemented in dProperties.

Specifically, a drug-like index equals 1 when logP (ALOGP), molar refractivity (AMR), molecular weight (MW), and number of atoms (nAT) of a compound are in the property range reported in the table below; moreover, the compound must be a combination of some of the following functional groups: a benzene ring, a heterocyclic ring (both aliphatic and aromatic), an aliphatic amine, a carboxamide group, an alcoholic hydroxyl group, a carboxy ester, and a keto group. For example, according to the CMC-80 index, an organic compound is a drug-like molecule if: the calculated ALOGP is between -0.4 and 5.6, the molar refractivity AMR between 40 and 130, the molecular weight MW between 160 and 480, the total number of atoms between 20 and 70, and it includes at least one of the above mentioned groups.

Two property ranges have been proposed: the qualifying range that covers approximately 80% of the drugs studied and the preferred range that is the smallest range within the qualifying range occupied by approximately 50% of the drugs. If large compound databases are screened by means of the indices based on the qualifying range (80%), the chance of missing drug-like compounds is less than 20%. To make the search/design for new drugs more efficient the indices based on the preferred range (50%) may be used, even if the chance of missing good compounds increases in this case.

Note that as these indices depend on ALOGP, their values are provided only for compounds having C, H, O, N, S, Se, P, B, Si, and halogens.