Interpretable machine learning methods for cross-omic mechanistic signatures of young hepatoblastoma
Dr. Claudio Angione of Teesside University discusses his research and why it is important for the future of treating childhood liver disease.
What is the study looking at?
This study aims to understand the way in which hepatoblastoma cells are influenced by their genetic make-up and by liver physiological conditions. Investigating cellular processes is done by the gathering ‘omic’ data. Omics refers to technologies used to explore the roles, relationships, and actions of various types of molecules that make up the cells of an organism. This data can be generated in large amounts by automated equipment.
Algorithms used in machine learning can dive into data in ways that humans can’t, detecting patterns that might otherwise be impossible to catch. However, one of the main weaknesses of machine learning techniques is that they do not incorporate biological knowledge in the learning process, and thus their predications can be difficult to interpret and trust. The analysis of such data is promising but doesn’t provide scientists with a deeper understanding of hepatoblastoma functioning.
What the study hopes to achieve
The combined use of machine learning methods (data-driven technology) and genome-scale metabolic models (knowledge-driven technology) will allow conclusions about the most likely outcome of the disease in each subject and the effect of potential drug interventions.
The methodology developed in this project may help to diagnose the disease earlier and suggest targets for novel drugs. Therefore, reducing the risks and failure rate of the treatment, where possible. Publicly available datasets of human hepatoblastoma from within the research community will be utilised.
Why is this research important?
Hepatoblastoma is the most common paediatric liver cancer, with an incidence that is increasing more quickly than for any other childhood cancer. It is a complex disease that cannot be fully explained by looking at single biological aspects, but requires a holistic approach bridging together genetic heredity, its regulation, and cellular metabolism. Early diagnosis, in general, allows choosing treatments with the highest rate of success and the smallest side effects based on the risk level of the tumour. However, establishing the risk level can often be difficult for clinicians. Furthermore, treatment of hepatoblastoma can be invasive and have several short or long-term side effects. Understanding the underlying mechanisms and functions of hepatoblastoma can provide more effective and less invasive treatments. In order to do this, it is necessary to analyse and interpret the large amount of data on liver cell activity through appropriate tools.
What about the future?
It is hoped that this combined method will have the potential to allow scientists to predict and explain biomarkers (a characteristic by which a particular disease can be identified) for hepatoblastoma with greater accuracy, leading to earlier diagnosis and more effective treatments.