Exploiting big data techniques to tackle child cancer
Treating cancer in children presents health professionals with a range of challenges. Cancer in children is relatively rare when compared to cancer in adults, and treatment options are often not always as efficient as one would hope. This is in part because cancer cells undergo many random changes (mutations). This results in each cancer having an essentially unique combination of molecular characteristics. “As a result, traditional treatment approaches are not always effective,” explains iPC project member Michelle Kölbl from Technikon in Austria. “This situation leads to a lack of targeted therapies, inaccurate risk level predictions, overtreatment and a compromised quality of life for young patients.”
Targeting individual cancer molecular profiles
To address this problem, the EU-funded iPC project sought to specifically tailor treatment combinations for the molecular profile of each individual cancer. The idea was that this would maximise cures, and minimise short- and long-term treatment side effects. In particular, the project team wanted to target the limited therapeutic options for children with paediatric cancers such as hepatoblastoma (HB), a cancer that forms in tissues around the liver. The team wanted to focus on patients with advanced tumours. What made the project unique was the development and application of machine learning and mechanistic models to predict optimal therapies for each individual child. To achieve its aims, the project pulled together a multidisciplinary team that included medical professionals and research institutes, as well as big data and computer modelling experts. “We used new computational and experimental approaches to analyse large-scale data sets,” adds Kölbl. “We measured experimental models and patient biopsies, to identify potential therapeutic avenues. These were then tested in cellular and animal models.” The project team looked at drug efficacy in HB patients, especially those with the aggressive C2 subtype associated with poor clinical outcomes. The approach relied on publicly available data.
Successful integration of advanced technologies
The project was able to demonstrate the effectiveness of using advanced computational models to systematically integrate vast quantities of data. The integration of advanced mechanistic, statistical and artificial intelligence models into a virtual patient framework enabled the team to carry out deep molecular analyses and treatment recommendations. “One of the key results from this work was the successful identification of drugs in rare paediatric cancers such as HB,” says Kölbl. “For example, this work led to the identification of cyclin-dependent kinase 9 (CDK9) inhibitors, called alvocidib and dinaciclib, as potent HB growth inhibitors for patients with the high-risk C2 molecular subtype.” Novel methodologies for personalised immune system modelling, automated comprehensive molecular tumour analyse reports, and a secure online platform for universal model access were other key milestones.
Bringing therapeutic approaches into clinical settings
Next steps include further validating these findings, and exploring how to bring identified therapeutic approaches into clinical settings. “The developed framework will empower clinicians with computational tools for personalised diagnostics and treatment recommendations,” notes Kölbl. “Over the long term, we hope that this project will demonstrate the usefulness of computational methods in identifying effective treatments for rare paediatric cancers, providing new therapeutic options, and potentially improving clinical outcomes for patients.”
Keywords
iPC, cancer, computational, drugs, molecular, hepatoblastoma, tumours, big data
