Class 13.

Towards a Systems View of Drug Discovery, Jordi Mestres

Apuntes Dabo aka El Calvo de Bogatell (versión con preposiciones)

Systems-blind drug discovery: you have a target and molecules that can bind to the
target.
Systems-aware drug discovery: you also take into account the ADME, diseases and side
effects, pathways and metabolites (they interfere in the organism with drugs)...
Side effects of drugs are the 4th cause of death.

Only 3000 proteins can be druggable (used as target), the bigger part are the kinases (22%).
Drug selectivity: there are a lot of molecules and the costs per molecule are extremely
high, so it is important to choose the compounds properly.
For each compound the number of receptors that are targeted is very high.
We have a limited capacity for screening. Selectivity is defined as our perception of the
ability of small molecules to bind to multiple proteins due to our limited capacity for screening
In the example Compound 1 is better than Compound 2 as the last one binds to more
receptors so it can be toxic:

Polypharmacology: the binding of a ligand to multiple protein targets. Not to be confused

with promiscuous aggregators due to self-association into colloids at high concentrations in
biological buffers. The pharmacological profile of a ligand is then characterized by the list of
binding targets and corresponding affinities.

The drug-target network is extremely complex, so a solution is to use computers. Using the
DrugBank database there are 1.7 targets per drug. Combining the DrugBank to Wombat we
have 2.7 targets per drug. Combining DrugBank, Wombat and Prediction databases there is
a mean of 6.3 targets per drug. This number of targets is important as apart from the
therapeutic effect we will have adverse effects that we want to minimize.
Antipsychotic drugs bind to everything. In the next Figure there is a comparison of the
promiscuity of the different compounds depending on their MW (molecular weight) and clogP
(related to hydrophobicity):

A drug is not selective for a target, one could be useful for the therapy but other could have
adverse effects.
Large drugs, polar, interact with a low number of proteins, however they bind to lots of
enzymes.
Just by looking at the molecule we can know where this molecule can bind.

Potential areas of impact of in silico target profiling:

The strategy of virtual screening is to use similar compounds, as they are more likely to bind
with the same proteins.
90% of drug candidates in clinical stage fail.

Drug repurposing: occurs when a drug is not toxic but not valid to its initial purpose, so you
try to find another purpose for this drug.

Conclusions: in silico target screening is emerging as a robust approach that allows to build
up on our understanding of how chemical entities interact with biological systems. The
potential impact of in silico target screening on both chemical biology and drug discovery will
hopefully increase the speed (and reduce dramatically the cost) at which new bioactive
chemical entities can be identified.
Biological systems are intrinsically robust: accordingly, selectively interacting with one
single target might not be the most efficient strategy to have therapeutic efficacy, as the
system may find other ways to compensate for the perturbation introduced. The alternative is
to interact with multiple targets, so making more difficult for the system to compensate for
all. The result is therapeutic efficacy but it does not come for free, as decompensation in
some particular systems may translate into adverse drug reactions.

A metabolomics view to personalised medicine:

1. A drug is an alien molecule that has been synthetically designed to interact optimally with
certain proteins (targets).
2. Most computational methods still treat drugs as isolated chemical entities.
3. This alien molecule will have to compete with endogenous metabolites evolutionary
designed to interact optimally with certain proteins as well.
4. The levels of endogenous metabolites vary across individuals.
5. Genetic and environmental factors can ultimately affect the levels of certain endogenous
metabolites.