Researchers Advance Cancer Drug-Matching, Personalized Medicine

A research team from Clemson University has created an open-source model designed to help medical researchers build scalable prediction models that can represent extensive cellular interactions and allow for the integration of large datasets, which could be used to advance personalized medicine approaches such as cancer drug-matching.

According to the press release, the study will be of most interest to researchers at pharmaceutical companies and those interested in computational modelling. These types of models are difficult to create for clinical applications because they are too small and limited in their capabilities to handle the large amounts of data required.

The researchers created their framework to address these issues and help others answer design questions, like how a manufacturer might run computer models of a product before actually building it.

“The impact of this paper is that we are trying to make it much, much easier to create these types of models,” said Cemal Erdem, postdoctoral fellow in Clemson’s department of chemical and biomolecular engineering and corresponding author on the paper. “We have this open-source tool now, with the code available on the internet. Researchers can take this code, create their own models and run simulations on their desktop computers or supercomputers.”

Because models to match drugs to patients are so difficult to build and test, there is very little foundation for researchers to start with.

“Medicine and pharma don’t have those types of design tools because they don’t exist yet,” noted Marc Birtwistle, an associate professor of chemical and biomolecular engineering at Clemson and corresponding author on the paper. “In a broad sense, that’s why this kind of work can be impactful. We’re trying to build that foundation so that those sorts of simulation models could help clinicians make decisions about patients.”

The model that the research team built for this purpose is intentionally simple and easy to use in an effort to promote accessibility and encourage other researchers to utilize it in their work. One pharmacy student has already begun using the platform, indicating that it is “very straightforward.” The press release states that her work with the model was underway within a matter of days, compared to the six months it would have taken without it.

Birtwistle stated that his overarching goal with the research is to match drugs to patients in the future.

This research is the latest in a series of efforts to use computational modelling and artificial intelligence (AI) to advance personalized and precision medicine.

Last month, researchers at Harvard and the University of Washington School of Medicine developed a deep learning tool capable of designing proteins that could be used in vaccines, medicines, and cancer treatments. Testing of the lab-generated proteins showed that they functioned as intended, with one successfully binding with anti-cancer receptors.