Drug discovery is complex and expensive. Numerous drug candidates fail late in
clinical trials or even after being released to the market. These failures are not only due to
commercial considerations and less optimal drug efficacies but, adverse reactions originating
from toxic effects also constitute a major challenge.
During the last two decades, significant advances have been made enabling the early prediction
of toxic effects using in silico techniques. However, by design, these essentially statistical
techniques have not taken the disease driving pathophysiological mechanisms into account.
The complexity of such mechanisms in combination with their interactions with drugspecific
properties and environmental and life-style related factors renders the task of predicting
toxicity on a purely statistical basis which is an insurmountable challenge. In response
to this situation, an interdisciplinary field has developed, referred to as systems toxicology,
where the notion of a network is used to integrate and model different types of information
to better predict drug toxicity. In this study, we briefly review the merits and limitations of such recent
promising predictive approaches integrating molecular networks, chemical compound networks, and protein drug