One of the main challenges around AI, in my view, is supporting an understanding of its application to real-world problems.

It is also something that may be holding back transforming the private and public sector through innovation, because organizations and individuals do not realize just how exciting and frankly essential AI, machine learning and data science are to our lives.

For example, how many people know what AI can do to successfully improve decision-making and efficiency in cancer treatments?

We all have a responsibility to communicate the work we are doing in the field of AI and machine learning. Whenever you mention cancer, this is something that is immediately relatable and comprehendible. Breast cancer is often detected after symptoms appear, while many women with breast cancer have no symptoms at all. Nevertheless, in both cases, cancer cells rewrite their ‘‘internal code’’ in order to satisfy the demand of growth and proliferation.

Such changes are driven by a combination of genetic (for example, gene mutations) and non-genetic factors (for example, tumor micro-environment) that result in an alteration of cellular metabolism. For this reason, any tests of early detection of breast cancer should be based on a joint analysis of both the genetic data and cellular metabolism of the patient.

Since early diagnosis and treatment are always associated with decreased breast cancer mortality, one of the greatest challenges in cancer research is to develop advanced methods for earlier detection. Nowadays, mammography screening is the most commonly used test for early breast cancer detection and it is associated with a 19% overall reduction in breast cancer mortality. However, several studies have analyzed the benefits and harms of this test showing that the net benefit of the screening strongly depends on a baseline breast cancer risk, which should be accounted for in the decision process. Indeed, mammography screening can also cause harm, particularly for younger women.

For this reason, guidelines recommend individualizing screening decisions depending on the age and clinical characteristics of the patient. This indicates a need to understand the relation between patient-specific features and cancer growth in order to develop personalized tests for earlier breast cancer detection. To develop such tests, it is necessary to predict which features have the greatest impact on cancer development. Many studies have identified the importance of metabolic reprogramming in cancer cells, giving metabolism a key role in cancer development. Indeed, the metabolism of cancer cells differs markedly from that of normal cells.

However, even if such behavior has been observed in many cancer types, the underlying consequences are still not wholly clarified, and an understanding of the causes of several cancer metabolic changes is lacking. Analyzing such metabolic changes between normal and cancer cells and identifying the factors that determine them could provide new insights in understanding cancer progression and detecting early breast cancer. Because of the high complexity of cancer metabolism, computational frameworks need to be developed for analyzing such a complex system: this is where AI comes in as a crucial application.

Since 2013, during my PhD at Cambridge University, I have worked on computational biology and machine learning applications in cancer to detect new cancer biomarkers and investigate genes networks. My publications include the development of a mathematical model to predict the formation of metastasis in breast cancer, a computational and statistical screening methodology to detect cancer biomarkers and predict survival probability, and optimization techniques to investigate metabolic changes in bacteria. All these research projects have a common theme: using computational and statistical methods to provide a better understanding of cancer progression and predict survival probability.

In my latest project with Dr Claudio Angione, leader of the Computational Systems Biology and Data Analytics research group at Teesside University, I am researching the application of AI-based computational models of breast cancer for early-detection personalized tests. Over the next two years, we will develop and validate a computational framework to investigate breast cancer metabolism. We will perform quantitative and predictive analysis of breast cancer data through survival analysis and machine learning techniques.

The analysis of patient-specific data is necessary to understand cancer development and identify a subset of features that will help towards early detection of cancer. In particular, we aim to develop a computational tool for clinicians that relates the profile of each patient to their probability of developing breast cancer based on a combination of machine learning and metabolic modeling techniques. This is just one example of a real-world application of AI and machine learning that could improve everyone’s lives.

Dr Annalisa Occhipinti is a lecturer in data analytics at Teesside University