Drug-drug interaction through molecular structure similarity analysis

Abstract

Background: Drug-drug interactions (DDIs) are responsible for many serious adverse events; their detection is crucial for patient safety but is very challenging. Currently, the US Food and Drug Administration and pharmaceutical companies are showing great interest in the development of improved tools for identifying DDIs.

Methods: We present a new methodology applicable on a large scale that identifies novel DDIs based on molecular structural similarity to drugs involved in established DDIs. The underlying assumption is that if drug A and drug B interact to produce a specific biological effect, then drugs similar to drug A (or drug B) are likely to interact with drug B (or drug A) to produce the same effect. DrugBank was used as a resource for collecting 9454 established DDIs. The structural similarity of all pairs of drugs in DrugBank was computed to identify DDI candidates.

Results: The methodology was evaluated using as a gold standard the interactions retrieved from the initial DrugBank database. Results demonstrated an overall sensitivity of 0.68, specificity of 0.96, and precision of 0.26. Additionally, the methodology was also evaluated in an independent test using the Micromedex/Drugdex database.

Conclusion: The proposed methodology is simple, efficient, allows the investigation of large numbers of drugs, and helps highlight the etiology of DDI. A database of 58 403 predicted DDIs with structural evidence is provided as an open resource for investigators seeking to analyze DDIs.

Figure 1

Overview of the construction of an interaction similarity model. Employing a list of known drug–drug interactions from DrugBank (step 1), structural similarity computation was carried out using molecular fingerprints (step 2) and a new list of predicted interactions based on structural similarity was generated (step 3).

Figure 2

Generating a drug–drug interactions (DDI) similarity model through combination of the DrugBank interaction database and molecular fingerprint-based modeling. In step 1, interaction matrix M₁ is created where the interactions in DrugBank are represented as ‘1’. In step 2, the similarity matrix M₂ is created based on the Tanimoto coefficient values. In step 3, M₁×M₂ is performed, the maximum value for each entry is retained, and the final matrix, M₃, is formed based on a symmetry-based transformation (retrieved interactions from M₁ when TC>0.75 are represented in red, new predicted interactions with TC>0.75 are represented in blue, and non-retrieved interactions from M₁ when TC>0.75 are represented in green). Values in the diagonal of all the matrices are 0 because the interaction of a drug with itself is not considered.

Figure 3

Receiver operating characteristic (ROC) curve showing the performance of the interaction model on the DrugBank database (430 128 possible interactions were computed). The area under the curve is 0.92.

Figure 4

Percentage of correct classifications for a random set of interactions described in Micromedex/Drugdex for the 50 most frequently sold drugs in 2009 (44 generic names) using the similarity interaction model (TC>0.75) and a random set of drug interactions (more details are provided in online supplementary tables S6 and S7). Drugs are sorted according to the percentage classified correctly by the model. Only interactions described in Micromedex/Drugdex but not in the DrugBank database are taken into account. % Correct classification (Micromedex/Drugdex interactions not described previously in DrugBank and correctly predicted by the model). % Correct classification (random set of interactions).