Predicting Drug-Drug Interaction Side Effects Using Label Propagation and Fire Hawk Optimization with Inceptionnet-Based Event Characterization

Drug-drug interactions, (DDIs) cause great worry in the medical community since they sometimes produce major side effects in patients. Understanding and expecting these side effects guarantees patient safety and helps one to maximize the effectiveness of treatment strategies. Although pharmacological research has advanced greatly, the complex interactions among drug-related entities still present difficulties for interpretation. To predict the negative effects of DDIs and so address this issue, we propose a new framework combining Fire Hawk Optimization (FHO) with Label Propagation (LP). Using the network-based similarity between drugs, label propagation finds possible interactions efficiently. On the other hand, Fire Hawk Optimization improves the representation of the complex interactions among drug-related entities by helping to extract feature interactions among them. Moreover applied as a predictive model is InceptionNet to evaluate and define DDI-related events. We leverage its deep hierarchical structure to extract extensive features. Included into a benchmark dataset for the tests were ten thousand known adverse drug reactions (DDIs), each consisting of 200 drugs and 50 potential side effects. With a sensitivity of 94.8% and a specificity of 97.1%, the proposed framework can thus achieve a prediction accuracy of 96.4%. The Fire Hawk Optimization method was able to significantly reduce the size of the feature set by 40% without compromising the accuracy of the prediction, even if the InceptionNet model attained a precision of 95.6%. These findings help to define the resilience of the system as well as its capacity to span a wide range of interaction models. This combination of LP, FHO, and InceptionNet has great potential for side effect prediction linked with DDI. Researchers and doctors have access to a consistent instrument that might improve drug safety profiles and reduce patient adverse effect count.