Drug Target Strategies

Network-driven drug discovery: Exploring biological and chemical space

• VF

  Victoria Flores, PhD, MSc

  Principal Scientist
  eTherapeutics

A characteristic of complex systems is that their behaviour emerges due to interactions between the multiple constituent entities, and therefore cannot be predicted by a linear combination of the behaviour of the individual components in isolation. This concept applies to biological systems where cellular functions arise from interactions between proteins, second messengers and metabolites within the cell. As such, cellular processes, including those manifested in different pathologies, can be modelled as a collection of interactions to capture their biological complexity. Interaction networks can be used as mathematical models to represent these complex systems.
Modelling cell complexity as networks to capture the changes that underpin disease could be of benefit to discover new therapeutic agents, as they should be better candidates to address biological degeneracy, cell robustness, as well as disease phenotypes arising from multi-component molecular changes. In the context of network biology, the drug discovery approach can be seen as the identification of agents that may have an effect on the disease by perturbing the underlying network.
We have implemented and used such approach in a drug discovery platform. Here, cellular disease mechanisms are modelled as protein interaction networks, and network theory-based algorithms are used to identify protein sets that upon perturbation, will disrupt the integrity of the disease network. The underlying assumption is that structural changes on the disease network translate into modifications on the disease phenotype.
The rationale for compound selection is based on their protein footprint being close to the protein set identified by network analysis. This compound selection step utilises empirical activity data from bioactivity databases together with activity predictions generated from machine learning models.
The computational process generates lists of compounds potentially enriched in actives for the selected disease mechanism. Those compounds are then tested in a variety of cell-based phenotypic assays representing the disease mechanisms being targeted. Hit compounds, defined as compounds active across the assay panel, are confirmed assessed for QC and and IP and taken into a medicinal chemistry programme for optimisation using phenotypic-driven approaches.
We have applied the platform described above for the discovery of novel drug candidates in diverse biologically complex diseases. The approach is highly productive and consistently identifies hits that have been progressed into novel potent and, selective drug molecule.