Using Heterogeneous Information Sources for Understanding the Mode of Action of Compounds

The mode of action of a compound in a biological system can be described on different levels, such as on the ligand-protein interaction level (e.g. binding to a target, inhibiting an enzyme etc.), as well as the effect level in a biological system (e.g. using genomic or proteomic readouts, cellular morphology readouts etc.). All of those viewpoints are equally valid, and hence an integration of different types of information to arrive at an integrated view on the mode of action of compounds in a biological system is highly desirable. This is where the current project aims to make a step forward. We will compile a compound data set which links molecular structure to bioactivites, and biological effects, on different levels, such as (but not limited to) the above. Subsequently, data mining algorithms will be employed that aim to establish links between the different levels of chemical and biological effects, i.e. to understand which interaction with which biological targets will lead to a particular downstream (often observable) effect. Within those links the next step is then to identify patterns, i.e. to be able to generalize which type of interaction, with which protein (or signalling cascade, etc.) will lead to a particular effect, which can then be translated to other, related biological systems (such as signalling networks). This work will be performed on both public and proprietary data, which will enable us to both publish our findings in a suitable manner, as well as to perform prospective validations with the company sponsor, AstraZeneca, in particular areas of interest.