Dutch researchers have built an artificial intelligence model that can predict adverse events associated with novel combination therapies for various types of cancer.
Clinical studies are the primary source of estimating drug-related adverse events. There is an increased interest in using drug combinations as they are more efficacious and avoid therapy-related resistance. However, there is no knowledge about possible adverse events due to different drug combinations.
The FDA adverse event reporting system (FAERS) database includes 15 million records of drug-related adverse events. Using this database, the Dutch researchers developed a method to predict drug combination-related adverse events and selected combinations with milder adverse event profiles. These findings were then used in a convolutional neural network algorithm — machine learning that mimics how human brains make associations between data.
They could generate an atlas that can accurately predict adverse event profiles for some of the most used combination therapies. However, the two main limitations of the study include difficulty comparing the data with more sparse data and a lack of clinical validation for the models’ findings.
The researchers are currently extracting adverse event-related data from the clinical records of the hospital. Additionally, they are linking the model to the data from the clinic. Parallelly, they are also building a prediction model to predict multitarget combinations based on clinical data. These data will help understand the side effects of combination therapies in the context of therapy efficacy.
Session OPO.BCS01.01 – Bioinformatics and Computational Biology 6312 – The adverse events atlas, towards a strategy to predict synergistic adverse events of combination therapies