Send to printer »

Feature Articles : Oct 15, 2010 (Vol. 30, No. 18)

Systems Biology Upends Drug Reprofiling

Rational Method Has Considerable Advantages over Shotgun Approach
  • Mireia Coma, Ph.D.
  • ,
  • Jordi Naval Chamosa, Ph.D.

There are smart thieves, and there are dumb thieves. Smart ones try to swiftly fit the right keys into the right locks, knowing which doors they wish to open. Dumb thieves try desperately to fit any kind of key into any kind of lock, no matter what door, and at the end they often get caught by the police. 

Following the same logic in the field of drug discovery, the right keys (compounds) must fit the right locks (targets) in order to open the right doors (cure diseases) and avoid the bad ones (adverse events). Unfortunately, the process of designing specific keys (compounds) for certain locks is becoming increasingly difficult in terms of cost-effectiveness.

As Nobel Laureate James W. Black once put it, “the most fruitful basis for the discovery of a new drug is to start with an old drug.” Accordingly, an attractive alternative to the constant search for successful drugs is drug reprofiling.

With reprofiling, the drug has often already undergone preclinical and clinical testing, thus reducing both the financial burden of having to conduct trials as well as the risk of adverse events in trial populations.

Drug candidates can be repositioned based on serendipitous observations, from novel, informed insights, or from technology platforms established to identify repositioning opportunities.

Several chemical libraries incorporate literature-based pharmacological data on key and lock affinities. These libraries connect keys (compounds) to locks (targets) on the basis of data published in scientific publications. They can be used for virtual affinity profiling to find the best key for a certain lock.

Among them, the WOMBAT database (Sunset Molecular Discovery) provides biological information for 309,847 compounds reported in medicinal chemistry journals or the MDL Drug Data Report, which includes information on therapeutic action and biological activity of more than 100,000 compounds.

Availability of keys is not often a limitation with such libraries. If the right “model” keys are available, i.e., compounds that have proven to be safe and effective for a given disease, there are methods to “mimic” those keys and find others that might be off-patent drugs, might be used as hits or starting points to pursue hit-to-lead refinement, or might be used in drug combinations, and thus qualify as new intellectual property.

Approaches based on computational chemistry and molecular modeling are commonly used by pharmaceutical companies to accelerate drug discovery and design. A number of companies provide receptor- or ligand-based virtual screening.

Also, public-private consortiums like the CTSA Pharmaceutical Assets Portal act as clearinghouses. Drugs discontinued at the clinical stage are made available by members of the consortium to be re-analyzed and re-entered in the translational research path.

There are a number of available compound libraries containing hundreds of thousands of compounds annotated to multiple targets. That means that one key can open multiple locks, and that those locks may or may not be related to each other (as the drug can bind to multiple active sites in the target), and also that we don’t know how many or which locks a certain key can open, and this lack of knowledge could lead to nice surprises (new indications) or to bad news (off-target effects).

The importance of target discovery necessitates the use of rational approaches in drug reprofiling and requires constant updating and refining of the key-to-lock strategy, both in database coverage and in specific one-to-one binding or locking simulations. This is especially important as new targets and new 3-D structures of those targets are constantly being added to the databases. 

From Targets to Disease

Let’s imagine that we have characterized the chemical structure of our molecule (consider it to be an old repurposed key), and we have defined a number of its targets. The next step is to identify which doors the locks may open. While it is desirable to open good doors (i.e., doors that have a positive therapeutic effect) there are also doors that we don’t want to open (i.e., doors harboring adverse events).

This is where a systems biology predictive in silico approach is essential. First, a complex map of protein-to-protein interactions, including signaling pathways, metabolic pathways, physical interactions between proteins, and known mechanisms of action, is created. These maps usually contain thousands of nodes (proteins), these nodes include the known targets and the known mechanisms of action of drugs.

Analysis of the map provides a number of insights, including discovering new therapeutic targets previously not identified; identifying new functions or therapeutic effects related to the known targets; and revealing new functions (new indications) of drugs that have multiple targets (i.e., opening multiple locks).

Map analysis can be facilitated with system biology tools, including mathematical network analysis, artificial intelligence, and genetic algorithms.

Use of systems biology approaches for reprofiling is game-changing for a number of reasons, chiefly that human physiology is complex, blurry, and redundant.

Successful approaches must model and simulate this complexity. In the pharmacological-chemical space, systems biology as applied to drug reprofiling is starting to do just that, and fast and efficient ways to rationally identify and predict the therapeutic performance of already known drugs are the result.