The development of computational models to predict the biological activity of small molecules has become an important part of modern drug discovery. Predictive modeling can support the process of drug discovery by helping medicinal chemists to select the most promising candidates for screenings or by providing valuable information about the underlying mechanisms of action. The latter, in particular, is of great interest for the development of new anti-infectives since biological assays for determining the potency of drug candidates often solely measure growth inhibition, which does not allow any conclusions to be drawn about the actual target hit.
The increasing availability of both comprehensive data and powerful algorithms has advanced research over the last few years and led to a variety of computational approaches for related tasks. But so far, only a few methods have been specifically developed and validated for microbial data. Further studies in this area are necessary not only because the development of new anti-infectives is in itself an important task, the general improvement of which is of great scientific and societal interest, but also because the underlying biological data have their own characteristics and limitations, so that findings from other areas cannot be transferred without validation. For example, looking at the availability of data, such as inhibitory concentrations of small molecules for bacterial and fungal targets, it is the case that significantly less data are available than for the more often well studied human targets. At the same time, the data used for modeling are potentially very diverse, considering the large number of existing assays and microorganisms for which a prediction of protein-drug interactions would be of interest.
The project focus is on validating ligand-based machine learning and data processing approaches, taking into account the aforementioned aspects and pursuing the overall objective of creating a tool which helps researchers assign mechanisms of action to drug candidates.
Name of Doctoral Researcher
Lennart Kinzel
Name of Supervisor
Prof. Dr. Knut Baumann
Institute / Department
Institute of Medicinal and Pharmaceutical Chemistry, TU Braunschweig
Contact details
l.kinzel@tu-braunschweig.de