Till date several predictive models have been
Till date, several predictive models have been developed by different research groups to speed up the process of library selection and drug optimization (Walters et al., 1999, Caldwell, 2000, Plewczynski et al., 2006, Wang and Ramnarayan, 1999, Pogorelcnik et al., 2015, Greenbaum et al., 2002, Sadowski, 2000, Charifson and Walters, 2000, Auer and Bajorath, 2006, Kumar and Zhang, 2016, Deng et al., 2006). Most of these methods rely on a piece of information gained through computational docking, ligand based pharmacophore or primary chemical attributes of a compound. Some of the previously described models are mentioned below:
Most of the above mentioned models require complicate statistical treatment, expensive computational resources, knowledge of programming or large number of empirical data. Considering all these problems, we wish to propose herein, a very simple predictive model for routine decision making and for CNQX classification. This method uses computationally derived efficiency indices in conjugation with comparative interaction profile of a compound with that of a reference. Although the usefulness of efficiency indices have been recently challenged (Kenny et al., 2014), but still a fairly large amount of literature support their utility specially at the initial stage of drug discovery (Cortes-Ciriano, 2016, Ponte-Sucre et al., 2015, Shultz, 2013, Abad-Zapatero and Metz, 2005, Schultes et al., 2010, Abad-Zapatero, 2007, Abad-Zapatero and Blasi, 2011, García-Sosa et al., 2011, García-Sosa et al., 2008, García-Sosa et al., 2010). Some of the facts in support of efficiency indices are as under:
By comparing the location of established drugs or known hits with the results of a particular high throughput virtual screening exercise (HTVS) in combined BEI-SEI plane (Abad-Zapatero, 2007) and in cluster dendrograms (Bouvier et al., 2010, Mantsyzov et al., 2012), several important trends can be noticed. Several important questions such as (1) whether our hits are lead/drug like or not, (2) in comparison of an established drugs what are the efficiencies and interaction profiles of our hits, (3) which kind of changes (such as introduction of polar group, change in the position of substitution etc) should be done to navigate to a particular direction in optimization plane, (4) which scaffold classes should be priorities in wet synthesis than others, can be answered (Abad-Zapatero, 2007, Abad-Zapatero and Blasi, 2011). In a way this type of exploration resembles SPR (structure property relationship) studies. Inclusion of several other indices and parameters related to physiology and toxicity (ADMET) can provide multidimensional framework to this study.
The above mention strategy can be easily applied to explore the chemico-biological space of several efficient reactions such as MCRs in structure base drug discovery (SBDD) efforts. Most of these reactions if not all, are quite efficient, atom and step economical, diversity oriented and suitable for automation (Teague et al., 1999, Ruijter et al., 2011). As a result, very large pool of diverse and biologically important scaffolds can be generated in a very shorter span of time. Wet lab synthesis and actual screening of all these compounds by conventional methods can be an expensive exercise, particularly when there is no previous history of these scaffolds against a known or novel target/(s). Computer based screening and insightful use of resultant data by methods, such as efficiency indices and pose clustering can be very useful in such instances and can save a lot of precious chemical, time and energy.
As an illustration we have navigated chemico-biological space of several annulated furanones against Pf-DHFR using the above mentioned idea of combined efficiency indices and pose clustering. There were primarily two reasons for the selection of annulated furanones:
Tools and techniques
Results and discussion Present study highlights the role of efficiency indices and pose-clustering in chemical prioritization and also provides a practical ground for selection of good starting point for drug optimization. In study like this, the final decision making largely depends on the capability of search algorithm, scoring function and on the reference values chosen. Hence in order to improve the predictive power of proposed model following factors were taken into account: