Drug-like properties of potential anti-cancer compounds from Cameroonian flora : A virtual studies

Article history: Received on: 11/01/2015 Revised on: 28/01/2015 Accepted on: 19/04/2015 Available online: 27/06/2015 Assessment of the ‘drug-like’ using a set of calculated molecular descriptors, of a collection of 195compounds from plants in Cameroon exhibiting in vitro or/and in vivo activities against various cancer cell lines was made. Lipinski’s ‘Rule of Five’ (Ro5) was used to evaluate drug-likeness/oral availability of the compounds by making use of popular parameters like Molecular Weight (MW), predicted lipophilicity (log P), number of Hydrogen Bond Donors/Acceptors (HBD and HBA) and Number of Rotatable Bonds (NRB). The drug-like properties of the studied compounds were compared to that of the Dictionary of Natural Products (DNP). Our results revealed that79.50%, 43.60% and 12.30% of our dataset fall within the recommended range for ‘drug-like’, ‘lead-like’ and ‘fragment-like’ respectively with significant enhancement of some oral availability parameters over DNP and this paints the significant of our dataset in search for lead-like anti-cancer molecules with desirable ‘drug-like’ properties.


INTRODUCTION
Cancer is a term used to refer to all malignant tumors.A neoplasm or tumor is an abnormal mass of tissue, the growth of which exceeds and is uncoordinated with that of the normal tissue and continues in the same manner after cessation of the stimuli which have initiated it (Ashutosh, 2007).Cancer is the second most common disease-related cause of death of human population (Gibbs, 2000).It is known to be responsible for about ~7.6 million deaths (~13% of all deaths) worldwide according to the World Health Organization (WHO, 2013).In spite of the enormous efforts and progress in the field of cancer research, the WHO estimates that the threatening of cancer disease will worsen (GLOBOCAN, 2008).Plant materials (and their derived photochemical) have been extensively used as chemotherapy to treat various forms of cancer and several reviews on medical plants used in the treatment of cancer and tumor have been published (Koehn et al., 2005;Hartwell, 1970;Graham et al., 2010;Rahman et al., 2013;Cragg and Newman, 2003).A report .by 1990 has it that about 80% of drugs were either Natural Products (NPs) or analogues inspired by them.According to a recent review on new chemical entities by Newman and Cragg, ~49% of anticancer drugs were either natural products or natural product-related synthetic compounds or their mimetics (Newman and Cragg, 2010).These data are coherent with a previous report by the same authors, which showed 47% of a total of 155 anticancer drugs approved up to 2006 were either natural products or directly derived from NPs lead compounds by semi-synthesis (Newman and Cragg, 2007).Cameroonian medicinal plants are endowed with NPs with intriguing chemical structures and promising biological activities (Pieme et al., 2012;Ndonsta et al., 2011).Many drugs with very interesting properties often fail to enter the market after much labor and investment, as a result of poor performance in physico-chemical assessment assays (Darvas et al., 2002), it has become customary to assess the 'drug-likeness' and physic-chemical properties of potential drug candidates early enough to cut down cost (DiMasi et al., 2003).This is often carried out by the use of experimental 'wet' screens or by computer-based (in silico) predictions using calculated molecular descriptors.The latter is much faster and less costly, making use of experimental data, which had accumulated by the close of the 1990s (Hodgson, 2001).
In silico methodologies make use of quantitative structure-activity relationships (QSARs) or knowledge-based approaches, using computer software (Button et al., 2003).In the current study, an attempt has been made to evaluate how 'drug-like' naturally occurring anticancer compounds are, by analyzing computed molecular descriptors for 195 anticancer compounds from Cameroonian medicinal plants and comparing them with that of DNP.This work is a follow up of our continuous development of NP libraries and in silico analysis of the 'drug-likeness' and physic-chemical profile, with the view of drug discovery (Ntie-Kang et al., 2013a;2013b;2013c;2013d;2013e;2014).

Data Sources
The chemical structures of pure compounds as well as their biological activities were retrieved from literature sources comprising mainly published articles from across the major journals of natural product chemistry.The exact chemical structures were confirmed from Dictionary of Natural Products.

Generation of 3D Models, Optimization and Calculation of Molecular Descriptors
Based on the known chemical structures of the NPs, all 3D molecular structures were generated using the graphical user interface (GUI) of the MOE software (Chemical Computing Group Inc., 2010)running on a Linux workstation with a 3.5 GHz Intel Core2 Duo processor.
The 3D structures were generated using the builder module of MOE and energy minimization was subsequently carried out using theMMFF94 force field (Halgren, 1996) until a gradient of 0.01 kcal/mol was reached.The 3D structures of the compounds were then saved as in pdb format included into a MOE database (.mdb) file which is suitable for use in several virtual screening workflow protocols.The Molar Weights (MW), Number Rotatable Bonds (NRB), lipophilicity (log P), Hydrogen Bond Acceptor (HBA), Hydrogen Bond Donor (HBD), and Lipinski violations were calculated using the molecular descriptor calculator included in the QuSAR module of the MOE package.

RESULTS AND DISCUSSION
The mean values for the relevant descriptors used to evaluate the 'drug-likeness' have been shown in table 1 for total compound library, as well as the standard 'drug-likeness', leadlike' and fragment-like' libraries.The concepts 'drug-likeness', lead-likeness' and 'fragment-likeness' have been respectively defined following the criteria for MW, log P, HBA, HBD and NRB: (MW<500, log P<5, HBD≤5, HBA≤10) (Lipinski et al., 1997); (150≤MW≥350, log P≤4, HBD≤3, HBA≤6) (Teague et al., 1999;Oprea, 2002;Schneider, 2002) and (MW≤250, -2≤log P≤3, HBD<3, HBA<6 and NRB<3) (Verdonk et al., 2003).This respectively gave 99, 66, and 23 compounds in the drug-like, lead-like and fragment-like subsets.Lipinski's test for oral availability Lipinski's 'rule of five' Ro5 has often been used as an assessment criterion for oral availability, although this rule sometimes does not apply to NPs (Lipinski, 2000;Quinn, 2008).It was observed that 78.97% of the studied compounds complied with the MW criterion (MW<500Da), while respectively 70.77%, 91.28% and 93.33% respected the log P<5, HBA≤10 and HBD≤5 parameters (Fig. 1).In addition, 72.82% of the compounds respected the criterion of the NRB, a rule often added to Lipinski's Ro5 to take drug metabolism and pharmacokinetics (DMPK) profiles into consideration.This globally showed that 155 of the compounds respected the Ro5, hence constituting the 'drug-like' subset.Meanwhile 72.82% of the compounds in the total library showed ≤2 violation of the Ro5 (Fig. 1F).The histogram of the variation of the MW parameter showed a peak between 300 and 400 Da (fig.1A).The computed log P values were within reasonable limits, with only values for five compounds exceeding 10 units (fig.1B).The log P distribution showed a rough Gaussian curve, with a peak centered at log P = 2 unit (fig.1B), while the HBA and HBD parameters both showed a rapid increase to peak values of 46 and 56 at four acceptors and three donor respectively (fig.1C and 1D).Both curves fell rapidly to 13 acceptors and 6 donors respectively for the bulkiest compounds.The curve of NRB (fig.1E) showed a peak values at zero rotatable bonds (RBs) with the bulkiest NPs having as many as 30 RBs.

Comparison with the Dictionary of Natural Products (DNP)
A comparison of the anticancer NP library has been carried out with the much larger Dictionary of Natural Products (DNP), containing 126,140 compounds (Dictionary of natural products, 2005).In our discussion, emphasis has been laid on the Lipinski compliance zones(MW<500, −2≤log P <5, HBA<10, and HBD<5), the values being expressed as percentage counts of the respective data sets.The distributions of the individual parameters for the two datasets have been shown in Fig. 3.These histograms generally show an enhancement of the distributions of the anticancer dataset over the DNP for Lipinski properties.The MW distribution histograms (Fig. 3A) showed that both curves peak at 301-400 Da.However, the anti-cancer dataset has higher percentage abundance for the regions301-400 and 401-500 Da within the Lipinski compliance zone with an enhancement of 9.20% and 2.33 % respectively in MW.Below this range, the percentages were reduced for our dataset, when compared to the DNP.The proportions of the two databases that satisfy Lipinski's MW property (<500 Da) were about 78.98 % for our dataset and lower (74.00 %) for the DNP dataset.The maximum values for the log P distributions were both at 2.5 log P units (Fig. 3B).The anticancer dataset is enhanced over the DNP for 1.0 < log P < 5 by 13.46% and a 25.70% enhancement of the DNP over the anticancer dataset for -2 < log P < 1.0.The HBA and HBD distributions (Fig. 3C-D) of the anti-cancer and DNP datasets respectively showed an interesting result.The peak of distributions for the HBD for the two datasets is situated at zero and two donors for DNP and anti-cancer datasets (22.50 % and 28.72%) respectively.Similarly, the peak of distributions for HBA for the two datasets is both situated at four at varying percentages of 14.15% and 23.59 % for DNP and anti-cancer datasets respectively.The overall summary of the four Lipinski parameters for the two datasets thus reveals that the anti-cancer dataset library is more 'drug-like' than the DNP, indicating that the chances of finding 'lead-like' molecules with improved DMPK properties within our library dataset are quite significant.

Fig. 3 :
Fig. 3:Comparison of property distribution for the two datasets by percentage distributions; A MW, B log P, C HBD and D HBA.For substructure B, the x-axis label is the lower limit of binned data, e.g.0 equivalents to 0 to 1.

Table 1 :
Summary of mean physic-chemical properties distributions of the library in comparison with the various subsets.