& PRESORT: Using computer modelling to find the best drug combination for each cancer patient
In the near future, oncologists may be able to use data from hundreds of genomic and proteomic analysis studies to make tailored therapies for treating their patients’ specific cancer. Currently though, experimentally searching these vast databases for effective drug combinations is far too expensive and time consuming to be therapeutically viable. Instead, researchers are borrowing techniques from engineering and computer modelling to develop a new field of systems biology to find potential cures for them.
“It’s a bit like trying to predict the weather”, says Dr Åsmund Flobak, lead researcher at DrugLogics, a partner project within the Centre for Digital Life Norway, describing the computer simulations he designed to understand how a cell will respond to treatment. “We have seen how many of the signalling mechanisms and networks work in other cells, and we are able to measure when some of the proteins are active or suppressed, but we don’t yet know how to put it together to predict which drugs will be effective”.
As an oncologist, Flobak was frustrated to see that drugs only helped some patients but not others and wanted to understand why. He decided to apply his knowledge of systems engineering to develop computer programs to solve this biological problem. In today’s world, nearly everything from medical devices to airplanes, are tested in computer simulations before coming to market. Flobak and DrugLogics are applying the same kind of modelling to cellular function to find the best cancer treatment for a patient.
Cellular simulations are only as good as the existing genomic and proteomic data they are based on. Dr Astrid Lægreid, the other half of DrugLogics’ leadership team, explains that one of the challenges to making DrugLogics widely applicable is incomplete databases or databases that include data from selected patient populations. The DrugLogics team is addressing these limitations by comparing the results from their models to experimental results from cancer cell lines and patient samples in the lab. Lægreid is working with the Norwegian Cancer Society to involve patients and caregivers to make sure that the results of the study are inclusive and are done in accordance with responsible research and innovation (RRI) best practices. The next step is validating those results on cell samples and verifying whether patients respond to therapies the same way.
To figure out how to implement these personalised cancer therapies, Flobak and Lægreid started the PRESORT project. They begin their research by biopsying live cancer cells from a cancer patient and identifying biomarkers. These biomarkers are used to calibrate computer models that propose a drug or combination of drugs for the patient. A high-speed robot then validates these top candidate drugs on the patient’s live cancer cells. Flobak and Lægreid are currently working with drugs for colorectal cancer but hope to use this model to identify new applications for these and other approved drugs. PRESORT’s long term goal is to create functionally validated system for fully individualised cancer therapy.
The PRESORT project: Tumour material is taken from patients via biopsy. Proteomic, transcriptomic and genomic biomarker data are fed into the DrugLogics modelling pipeline, along with prior knowledge on cancer signalling. Machine learning algorithms optimise computational models. Meanwhile, the patient cancer cells are expanded, and then undergo testing with a limited number of computationally predicted effective drugs. Results can be used for clinical decision support.
Successfully developing the models in DrugLogics and PRESORT and bringing them to patients will require joint global commitment. The databases need to improve, but a similar proteomic database and input from pharmaceutical companies are also needed. Modelling the whole cell, the whole tumour, and ultimately the whole body, to design tailored treatments is a challenge but will be immensely beneficial to people with colon cancer, inflammatory disease, and potentially even the next pandemic. “In the end we must do better than weather predictions, and given how hard predicting the weather is, this calls for large research collaborations”, says Flobak.