| Drug development is currently a time-consuming, costly and challenging process. The process typically starts with the identification of a therapeutic target for a given disease. A therapeutic target is some biological molecule and the binding of compounds to target molecules is expected to cause a desired therapeutic effect. That is, target binding compounds have the potential to become drug candidates. However, there is a tendency for many drug candidates to fail during clinical trials, and consequently, very few candidates become approved new drugs. This trend suggests that the early stages of drug development should be improved to provide better drug candidates.;Overall, this thesis shows how computational prediction in a systems biology framework can be used to facilitate and expedite the early stages of drug development.;The reasons for which a drug candidate may fail during clinical trials include unacceptable toxicity and insufficient efficacy observed in humans. These reasons suggest that the assessments of a compound during the early stages of drug development often inaccurately predict the effect of the compound in humans. One of the main goals of systems biology is to accurately predict how a given biological system responds to perturbations, e.g. treatment with a compound. This suggests that systems biology can help address challenges in drug development. However, there are currently gaps in our knowledge of systems. Here we use machine learning techniques to exploit existing systems data towards filling in these gaps. In particular, we developed a method that uses the occurrences of motifs in protein sequences to predict kinase-substrate interactions. We also developed a method that uses gene expression, protein-protein interaction and phenotype data to predict genetic interactions. These predicted interactions can facilitate the identification of potential therapeutic targets. Ultimately, a better selection of therapeutic targets should lead to better drug candidates. We also address the challenge of developing combinatorial therapies. Despite the fact that combinatorial therapies are advantageous, the scale of the experiments required to search for desirable chemical combinations is currently prohibitive. We therefore developed a method that uses system response data to predict chemical synergies towards facilitating the development of combinatorial therapies. |
