There is a growing perception that the traditional approach for drug development has not been as effective as could be expected [1
]. Although investments by pharmaceutical companies have been growing continuously in the last decades, the number of newly approved drugs has not followed the same trend [2
]. Currently the main trend in drug design and development follows the target-based approach. The drug target is generally a single gene or gene product which has been clearly identified as having an action on the disease. Disease treatment seeks to modulate this action without affecting other processes in the organism. However, many diseases are multifactorial, and the current approach fails to take such systemic aspects into account [4
]. The complex interactions between numerous molecular processes and pathways involved in diseases imply that any treatment targeting a local element will create secondary effects on the entire system. The lack of an integrative investigation of drug action can lead to either positive or negative systemic effects being missed. In the first case, the treatment only provides limited and localized cure for the disease while failing to address large-scale causes. In the second case, new drugs are developed with harmful side effects that are only revealed at a late stage of clinical studies. This problem of drug attrition is a major concern for the pharmaceutical industry and an important source of financial costs. Additional problems can be caused by side-effects resulting from combinations of several drugs, arising from unexpected interactions which were not revealed by local analysis.
To increase the productivity of drug development and the efficiency of new drugs, there is thus a strong need to move beyond the investigation of drug action from the point of view of direct targets only, and to take into account the full network of interactions connecting the targets with the rest of the system. Drug development needs to be approached not only in terms of a molecular biology process, but also as a systems biology process [7
]. This awareness has recently led to the completion of major studies revealing the large-scale interactions between drugs, drug targets and diseases from a network-based perspective [11
]. However, these approaches did not unveil yet how specific drugs interact in biochemical systems, and new methods are needed to gain an integrated perspective of the mechanisms of drug action in complex systems.
To this aim, we here introduce a new view of drug-metabolism interactions based on the concept of metabolic drug scope
. The concept of a scope
was previously developed in metabolic networks [16
]. It uses an expansion process based on the principle that, for any reaction to take place, all necessary substrates must be present and the products of these reactions may in turn be used by other reactions. Starting from a set of seed compounds, reactions whose substrates are available in the seed set are iteratively added, resulting in the generation of a series of expanding networks. When no further reaction can be added, the final network obtained is defined as the scope
of the seeds. The scope represents the set of all compounds that can in principle be synthesized from the seeds. The properties of scopes in metabolic networks have been described in great detail [17
]. This concept was already proven valuable in revealing features of the evolution of metabolic systems, leading notably to demonstrating how the shift from anoxic to oxic environments transformed the architecture of metabolic networks and how the availability of oxygen allowed an increase in biological complexity [20
]. Recently, it was shown that features of scopes could also be linked to chemical structures and biological functions of the seeds [21
We here extend the scope paradigm to drug-metabolism interactions. Many drugs target genes or gene products that are involved in metabolic functions. When a metabolic reaction is targeted by a drug, the concentrations of its substrates and products are affected as a result of drug action. Other reactions which use the products of the initial reactions are in turn affected by these perturbations. By iterative expansion, the entire scope of the initial substrates and products can in principle be affected by the drug. Basing on this principle, we thus introduce the concept of a metabolic drug scope defined by the following: the scope of a drug is the scope resulting from the expansion of a set of seed compounds containing the substrates and products of all metabolic reactions targeted by that drug.
The metabolic drug scope essentially represents the largest network of possible action of a drug in a metabolic system. In the following sections, we show how metabolic drug scopes reveal new features of the systemic modes of drug action through a detailed analysis of the scopes of 276 human approved drugs from the DrugBank database [22
] having metabolic targets. As scopes depend on the topology of the metabolic network used in the expansion process, we analyzed the action of each drug in four different systems: a human vs. a reference metabolic network consisting of the union of all organisms, and for each of those a reversible vs. an irreversible topology. Further details on the construction of metabolic drug scopes are provided in the Methods section. Our results show that the distribution of scopes is highly uneven and that they can be classified into different categories. Some drugs have small scopes associated to localized action, while others have large scopes associated to potential widespread systemic action. These classes are relatively well conserved throughout the four systems. A network of metabolic drug scopes, where scopes are connected when their Jaccard distance is smaller than a given threshold, confirms these characteristics, and a k-core decomposition algorithm reveals several highly connected central components as well as peripheral subgraphs. We subsequently analyzed the correlations between metabolic drug scopes and therapeutic properties of drugs, and show that therapies are not distributed uniformly but can often associated to specific groups of scopes, correlated to similar k-cores. Metabolic drug scopes therefore offer a new avenue to analyze drug-metabolism interactions and may provide valuable assistance to the drug development and assessment process.