Enhanced prediction of drug interactions using deep learning in drug-drug and drug-food interactions expanded to interaction of Paxlovid

Abstract

Drug interactions including drug–drug interactions (DDIs) and drug–food constituent interactions (DFIs) can trigger unexpected pharmacological effects such as adverse drug events (ADEs). In this respect, drug interactions with urgently approved drugs can be a serious medical problem for COVID-19 patients who have been prescribed medicines of chronic diseases. Pfizer’s Paxlovid is representative approved drug for the treatment of COVID-19. Here, we improved the previously developed computational framework DeepDDI to provide sufficient details of DDI, and DeepDDI 2 can predict 113 types of drug interactions beyond 86 labels which previous DeepDDI provided. A structural similarity profile (SSP) for an drug was generated by obtaining Tanimoto coefficient between an input drug and 2,386 approved drugs in a pairwise manner. Then, SSPs of each drug in an input pair were merged into a feature vector for DeepDDI 2 training. The structure of DeepDDI 2 is a multilayer perceptron which has 113 output neurons where each represents a specific DDI type. DeepDDI 2 predicts potential drug–drug interactions between Paxlovid components (nirmatrelvir and ritonavir) and 2,248 prescription drugs. As a result, nirmatrelvir is expected to interact with 673 approved drugs and ritonavir with 1,403 approved drugs. In addition, we proposed 239 and 124 alternative drugs that would reduce the likelihood of drug interaction with nirmatrelvir or ritonavir, respectively. The updated DeepDDI 2 provides not only useful information in drug prescription, but also shows the possibility of AI in digital healthcare.