ABSTRACT
The compound-target matrix plays a central role in chemogenomics. Its columns are formed by the set of all possible targets encoded in the genes of organisms (not necessarily only human genes), and the rows represent all the compounds that span the huge chemical space of fragments and lead-or drug-like compounds. The matrix elements describe the biological interaction, for example, a classiication as active/inactive or a quantitative description by IC50/EC50 or raw % CTRL values. Each row of this matrix displays the activity proile (the bioprint) of a compound, and each column displays the compound-binding proile of a target (the chemoprint). Regarding experimental data the compound-target matrix is and will remain extremely sparse. Given the huge size of the relevant chemical space and ten thousands of potential targets, it is obviously impossible to ill the matrix with assay data. Hence, in silico approaches are the alternative to complement the bioprints of the compounds and the chemoprints of the targets. Calculating the interaction strength of a wide diversity of compounds and targets represents a challenging goal, and computational chemogenomics is by far not yet mature enough to always provide reliable predictions. Despite this, it is a very attractive goal for pharmaceutical research. The prediction of the biological proile of compounds would allow the identiication of potential off-targets, which may cause unwanted side effects of a drug. This information would help to prioritize the targets for the safety proiling and could be used to optimize compounds toward reduced side effects. Knowledge of the similarity between proteins can pave the way to chemical starting points or tools for innovative targets. There is also increasing evidence that most if not all drugs bind to a variety of targets (called polypharmacology) with relevance for the therapeutic action of the drugs and/or for the side effects [2]. The knowledge of the target spectrum of a drug is crucial information for the so-called drug repurposing where a known drug is applied to a new disease. It can also help to get better insight into the diseaserelevant targets and pathways and to identify new and better approaches to treat a disease, for example, by multitarget drugs [3, 4]. Moreover in phenotypic screening the target is mostly unknown and the activity proile of an active compound may be the key to the identiication of the relevant target(s) [5].
