Note: Maximizing computational tools for successful drug discovery

Note for: Maximizing computational tools for successful drug discovery

(doi: 10.1517/17460441.2015.1016497)

Selecting the appropriate computational tools is dependent on the goal of the drug discovery project - for instance -

1. understanding the SAR of a compound series

2. synthesizing/optimizing novel leads

3. understanding the binding modality for compounds of interest

4. screening for novel compounds with desired bioactivity

5. identifying potential toxicity in compounds

6. designing multi-target inhibitors

7. predicting the side effects of drugs

Aim of this review: this editorial will briefly touch upon several key steps in the drug discovery process and their associated tools in the context of trying to cover most (if not all) of these essential topics in drug discovery.

we also confer known or potential opportunities and problems that researchers may need to be aware of in their drug discovery endeavors.

Despite the seemingly large chemical library produced by these efforts, we are still far from reaching the estimated chemical space of > 10^60 molecules.

Scaffold hopping is to discover structurally novel compounds starting from known active compounds by modifying the central core structure of the molecule (10.1016/j.ddtec.2004.10.009).

Curation of these bioactivity data is available in public databases such as PubChem, ChEMBL, and BindingDB.

In spite of growing progress in finding cures for several diseases, the lack of treatment and bioactivity data still persists for rare and neglected diseases.

Bayer HealthCare initiated a crowdsourcing project called Grants4Targets that aims to foster collaboration between researchers in academia and pharmaceutical

industry.

Virtual screening can be classified into two major types:

i) ligand-based virtual screening; 

ii) structure-based (or target-based) virtual screening.

Ligand-based virtual screening -- based on the assumption that compounds with similar chemical structures have the tendency to share similar biology activities.

QSAR in drug discovery generally fall into the following categories: prediction of ADMET properties, drug-likeness profiling, as well as predicting inhibition of target proteins], cancer cells and pathogens. the protein structure for investigating characteristics of the binding pocket, de novo design of ligands inside the binding pocket and performing protein-ligand docking.

The protein structure for investigating characteristics of the binding pocket, de novo design of ligands inside the binding

pocket and performing protein-ligand docking.

Molecular dynamics (MD) has been shown to improve the performance of virtual screening against known crystal structures as it can generate snapshots for subsequent sampling of relevant protein conformation.

Dock several proteins to the same ligand (an approach called reverse or inverse docking) as well as cross-docking several ligands against an ensemble of proteins.

                                                                                             

If crystal structures are available;

1. AutoGrow

2. LigMerge

These two tools provide the knowledge on ligand design that could fit well to the structure and affecting the activity.

Notable tools for performing shape and charge similarity search are ROCS and EON, respectively, both available from OpenEye (http://www.eyesopen.com/). Although both are commercial software but an academic license is also available for researchers from academia.

Screen3D has recently been reported by ChemAxon for performing shape similarity search.

ZINC addresses this issue as it provides access to 35 million compounds that are also commercially available.

ChemSpider is another equally important repository that aside from providing access to information on 32 million compounds not only links to chemical vendors but also links to patents, physicochemical information and other relevant repositories.

virtual screening was classified into four major classes (i.e., classical, parallel, iterative and integrated) as functions of how wet and virtual screening are connected.

Potential pitfalls of virtual screening;

1. erroneous assumptions and expectations

2. data design and content

3. conformational sampling and

ligand/target flexibility

4. choice of software

redundancy and resiliency in biological networks that are commonly found in complex diseases (i.e., cancer, diabetes and cardiovascular diseases).

Once an undesirable property of the single-target paradigm (sometimes referred to as off-target binding, binding promiscuity or polypharmacology), multi-target drugs have found a new therapeutic route in its rebranding as drug repositioning (also called drug repurposing) in which existing FDA-approved drugs are reused for treating another disease.

Proteochemometrics is instrumental in discerning the structure--activity relationship for a series of compound against a series of proteins and it has been successfully applied on a wide range of protein families such as G-protein-coupled receptors, proteases, kinases and cytochrome P450s among others.

Reverse or inverse docking modifies the original concept of docking a series of ligands against a single-target protein to the docking of a series of proteins to a target compound.

Aiming to tackle this gap (large data on ligands but few info on the protein structure) is the CARLSBAD database that houses nearly 1,500,000 unique bioactivities against > 400,000 compounds. Similarly, AstraZeneca developed the SAR Connect [52] for mining the big data in drug discovery compiled from ChEMBL, GOSTAR and AstraZeneca’s IBIS.

The benefits of allosteric drugs is its greater specificity, reduced side effects and lower toxicity as well as its usage flexibility in which it can be used to target large protein--protein interaction for which small molecules may not be applicable as well as used to increase activity of its target and tackle drug-resistant targets.

AlloSteric Database [57] had recently updated its repository to encompass 1286 allosteric proteins and 22,008 allosteric modulators.

The lack of consistency and interoperability of computational tools presents a major bottleneck in drug discovery.

Moreover, Cinfony establishes itself as a central platform of cheminformatics toolkits by housing them all in one common interface as a Python module.

Furthermore, data from biological databases (i.e., BioModels, ChEMBL, KEGG and UniProt) can be programmatically accessed via a common Python framework called BioServices.

Ongoing efforts on achieving interoperability via establishment of open standards and open source software have been carried out by the Blue Obelisk and (Open PHArmacological Concepts Triple Store) consortiums.

Selection of appropriate computational tools is highly dependent on the available data at hand and on a case-by-case basis.

For example, if available data are primarily based on known active ligands in the absence of the protein crystal structure then employing ligand-based virtual screening approaches may be preferable unless a suitable homologous protein can be identified that can serve as a template for protein structure prediction thereby opening up the possibility to use structure-based approaches.

obtaining compounds with good pharmacokinetic profiles and low drug adverse effects.

it is necessary to develop a more streamlined infrastructure that better supports interoperability among various databases and tools.