Review of Predicting Synergistic Drug Combinations

Abstract

The prediction of drug combinations is of great clinical significance. In many diseases, such as high blood pressure, diabetes, and stomach ulcers, the simultaneous use of two or more drugs has shown clear efficacy. It has greatly reduced the progression of drug resistance. This review presents the latest applications of methods for predicting the effects of drug combinations and the bioactivity databases commonly used in drug combination prediction. These studies have played a significant role in developing precision therapy. We first describe the concept of synergy. we study various publicly available databases for drug combination prediction tasks. Next, we introduce five algorithms applied to drug combinatorial prediction, which include traditional machine learning methods, deep learning methods, mathematical methods, systems biology methods and search algorithms. In the end, we sum up the difficulties encountered in prediction models.

HANSynergy: Heterogeneous Graph Attention Network for Drug Synergy Prediction

“…Nevertheless, greater attention should be paid to the biological connections between drugs and cell lines. 24 Therefore, the construction of heterogeneous biological information networks as a powerful tool and method for further study of drug and cell line connections has been an emerging research direction in the field of bioinformatics. This approach centers on the amalgamation of varied biological data sources into an extensive network graph, aiming to attain a deeper comprehension of the intricate nature of biological systems.…”

“…Computational approaches, including algorithms based on machine learning − and network analysis, − play a pivotal role in handling extensive data sets and discerning meaningful patterns. These approaches integrate information from various sources, including molecular structures, cellular pathways, and clinical data, to create predictive models. , By analyzing the interactions between drugs at the molecular level, researchers can anticipate potential effects and identify combinations that may enhance therapeutic effects . In recent years, alongside the enhancement of computing power resources and the progress of artificial intelligence, deep learning (DL) has proved to be superior to the manual feature extraction methods of traditional machine learning .…”

“…These approaches deviated from traditional methods that obtain drug combination features by concatenating individual drug features. Nevertheless, greater attention should be paid to the biological connections between drugs and cell lines …”

Environment by environment interactions (ExE) differ across genetic backgrounds (ExExG)

“…But just as genotype-phenotype mapping studies focus less on environment interactions, drug synergy studies focus less on genetic interactions. For example, several studies suggest that if one understands the cell biology underlying drug interactions, one can predict synergy (Feala et al, 2010; Pan et al, 2023; Yang et al, 2020), but this ignores that mutations may change the underlying drug interactions (Schmidlin et al, 2024). Other studies describe the biggest challenge in detecting synergy as there being more possible drug combinations than one can study (Madani Tonekaboni et al, 2018; Menden et al, 2019; Pan et al, 2023), but this ignores that studying every drug combination in every genetic background would be orders of magnitude more difficult.…”

“…For example, several studies suggest that if one understands the cell biology underlying drug interactions, one can predict synergy (Feala et al, 2010; Pan et al, 2023; Yang et al, 2020), but this ignores that mutations may change the underlying drug interactions (Schmidlin et al, 2024). Other studies describe the biggest challenge in detecting synergy as there being more possible drug combinations than one can study (Madani Tonekaboni et al, 2018; Menden et al, 2019; Pan et al, 2023), but this ignores that studying every drug combination in every genetic background would be orders of magnitude more difficult. Despite the combinatorics challenge, heroic efforts have been made to measure massive numbers of drug interactions (Menden et al, 2019), including higher-order interactions (Lozano-Huntelman et al, 2021), which have fueled sophisticated multidrug treatment strategies and evolutionary models (Baym et al, 2016).…”