Unlocking the future of personalised cancer treatment with the help of AI
Conventional cancer therapy using chemotherapeutic drugs tends to treat patients similarly with one-size-fits-all approaches. When these treatments stop working, as for relapsed or refractory patients, clinicians are left with limited treatment alternatives. In most cases, there is no way to predict whether the patients will respond or not. Similarly, pharmaceutical companies often face discouraging outcomes in clinical trials because of their limited understanding of the factors that determine patient response. Clinical phenotypes or genomics signatures are employed for patient stratification, but these do not necessarily predict drug responses. Therefore, reliable drug sensitivity prediction tools are needed to help clinicians and pharmaceutical companies bring more effective drugs to cancer patients. Moreover, cancer cells have the propensity to develop resistance to drugs, necessitating the use of drug combinations for good clinical outcomes. However, identifying synergistic and effective drug combinations has traditionally been a time-consuming and often unpredictable process.
Computational tools for personalised cancer therapy
Funded by the European Research Council, the DrugComb project developed mathematical and computational tools to identify and prioritise effective drug combinations for cancer patients. Additionally, it undertook drug combination screening studies on cancer cells for experimental validation. “Our goal was to understand the potential synergy between drugs and translate the knowledge into treatment suggestions for cancer patients,” explains project coordinator Jing Tang.
Applying AI tools in predicting and testing drug interactions
Researchers employed AI-based text mining techniques to extract experimental drug-target interaction data from scientific literature. Drug sensitivity and drug target data was then incorporated into the DrugComb data portal that includes tools for the analysis of drug combinations. The project also developed an open-source tool called SynergyFinder Plus (www.synergyfinder.org) to investigate the synergy of approved and clinical trial drugs in patient-extracted cancer samples. If a drug combination was synergistic against cancer cells but not damaging to normal cells, then the drug combination was considered as a potential candidate. Furthermore, AI-based algorithms allowed scientists to discover the links between drug synergies and cancer driver genes, for improved treatment efficacy with fewer side effects.
Added value of AI in predicting drug combinations
Personalised medicine as it is practiced today usually does not outline the unique characteristics of individual patients. Furthermore, most of the treatment options involve only single drugs, whose efficacy is limited due to the emergence of drug resistance. “The DrugComb computational analysis pipeline has exceptional potential to lead to novel, more effective and safe treatments compared to the current cytotoxic monotherapies,” emphasises Tang. DrugComb constitutes one of the pioneers in computational tool development for drug combination predictions. The main novelty in the DrugComb solution lies in the fact that it combines genomics with drug sensitivity data to provide more complete and dynamic profiles about disease in real time. The platform has been successfully used to predict synergistic combinations in patients with T-cell prolymphocytic leukaemia and breast cancer. Since completing the project, multiple pharmaceutical companies have expressed interest in licensing of the DrugComb portal for data analysis. The team is considering commercial versions of the software tools for pharmaceutical companies. Tang envisions “a company focused on AI-assisted drug discovery, utilising the databases and tools developed by DrugComb for increased cost-efficiency and improved patient care.”
Keywords
DrugComb, cancer, AI, drug sensitivity, pharmaceutical, computational tool, genomics, drug response, personalised medicine
