Prediction of Effective Drug Combinations based on Potential Drug Profiles
Abstract
Cancer is a great threat to the health of all mankind, and cancer monotherapy has been characterized by drawbacks such as toxicity and drug resistance. With the development of network pharmacology, multi-targeted drug combinations have become an ideal choice for cancer treatment. The dosage of combination drugs is usually lower than that of monotherapies, which has the advantages of improving efficacy, reducing toxicity, and delaying the development of drug resistance. In order to obtain better prediction results, this paper proposes a method for constructing drug potential features based on graph embedding model to predict anticancer drug combinations, establishes a control group to validate our method, and selects four performance metrics to measure the prediction performance of the model. The results show that the prediction results obtained from the drug potential features are better than the drug features. The drug potential features we designed can be used as one of the optional features for predicting drug combinations.
Keywords
Download Options
Introduction
Chemotherapy is a commonly used treatment for cancer, which often has many side effects, such as drug resistance and toxicity. With the development of modern medicine, drug combinations have become an ideal method for cancer treatment. By combining two or more anticancer drugs, drug toxicity can be reduced, drug resistance can be delayed, and efficacy can be improved. Therefore, finding synergistic combinations of drugs for specific cancer types is important to improve the efficacy of anticancer therapy[1-3].
Methods such as machine learning models offer the possibility to explore the combination space effectively. Machine learning models can quickly adapt to the ever-changing task of anticancer drug combination prediction and continuously optimize the prediction results. For example, by using machine learning models such as support vector machines, random forests, and neural networks, synergistic effects between different drugs can be effectively predicted[2, 4, 5].
Conclusion
The use of combinatorial drugs can undoubtedly help people to treat complex diseases, while the use of computer technology, histology and network technology helps people to discover new drug combinations and is therefore a proven means. The method greatly reduces the scope of the search and is safer and more reliable in a small area before experimental tests are carried out. Discovering new reliable features is also one of the keys to accurate prediction, and the drug potential features designed in this paper play an important role in improving the prediction accuracy, and among the four models, the contribution of drug potential features to the prediction accuracy is significantly higher than that of drug features. The Morgan molecular fingerprints, drug target information and monotherapy information used in this paper are basically classical and have been used by previous authors, so their credibility can be guaranteed.
The principle of constructing drug potential features in this paper lies in the need for a large amount of drug synergy information, and the accuracy can be further increased if enough drug synergy information is available. In addition, the drug potential features constructed in this paper can also be utilized in the migration learning method, using a large number of drugdrug interactions in the data set, to extract the potential information to construct drug potential features, which can be applied to other prediction tasks that lack cell line information or drug feature information, which is also one of the future research directions. Finally, the models used in this paper are supervised models, and it is believed that the prediction accuracy will be further improved if semi-supervised models or other more advanced algorithms are utilized.