Acute Myeloid Leukaemia by Personalised Medicine – improved treatments of acute myeloid leukaemias by personalised medicine.
Project lead: Inge Jonassen
Institution: University of Bergen (UiB)
Partners: Haukeland University Hospital, Norway; University of Groningen, The Netherlands; German Cancer Research Center (DKFZ), Germany; University of Freiburg, Germany; Medical Center Leeuwarden, The Netherlands; University of Toronto, Canada
Funding: ERAPerMed (Research Council of Norway)
Duration: 3 years
Acute myeloid leukaemia (AML) is a fast-acting cancer that can be treated, but the drugs are harsh and the patients, many of whom are elderly, are often too weak to tolerate many attempts at finding the right one. Patients are much more likely to become long-term survivors if their doctor’s first attempt at therapy is highly effective. Therefore, it is essential to have tools to determine which treatment will work best for the patient the first time. While many labs are developing personalised medicine schemes for patients based on genomic data, the Acute myeloid leukaemia by personalised medicine research group is using machine learning to find the right treatment based on analysing how the signalling pathways in the patient’s cells affect how the disease develops.
AML is a type of cancer that impairs the body’s ability to correctly make blood cells, resulting in abnormal blood cells building up in the body and damaging other organs. Leukaemia is a difficult disease to study and treat because it has many different causes. In each case though, there is a problem with the signalling molecules that turn on and off the genetic programming that transforms bone marrow stem cells into mature blood cells. Researchers from fields as diverse as bioinformatics, cell biology, and machine learning are studying how these signalling molecules function, but it is impossible to directly measure them.
Using computational biology, the researchers in Acute myeloid leukaemia by personalised medicine are looking for patterns in leukaemia cells that connect the signalling molecules to the abnormal cell types. They can use this information to model different disease progressions and treatments. So far, they have tested over 100,000 cell samples for gene expression and signalling molecules. They still need to do animal and clinical trials to test their machine learning algorithm, but they ultimately imagine having a diagnostic tool that could measure the genomic and signalling pathway of a patient with leukaemia and predict which treatments will be effective before trying it on the patient.
The researchers are currently training their algorithm on consistent samples of a small sub-type of leukaemia, but expect their model to be applicable to all types of leukaemia. Their published work and software will be open source and available to the public to use so that their unique approach of focusing on signalling pathways in systems biology can contribute to the global research effort to develop personalised medicine for all diseases.