Research Spotlight: AI-driven clinical translation of AFM biomarkers for cancer diagnosis

A key focus of recent research led by Conway Fellows, Prof. Brian Rodriguez from UCD School of Physics and Dr Stephen Thorpe from UCD School of Medicine is the application of artificial intelligence (AI) in the analysis of tissue and cell mechanics, particularly in the context of cancer diagnosis and prognosis.

The team recently published a review that explores the current state of AI applications, particularly in the use of atomic force microscopy (AFM) to assess tumour stiffness.

Solid tumours are generally much stiffer than the normal tissue they originate from. This increased stiffness can significantly contribute to treatment resistance in several ways.

Tumours that are stiffer typically have a dense structure. This is often due to the presence of collagen that can make up a large portion of the tumour mass. This dense extracellular matrix creates a physical barrier that can obstruct the delivery of therapeutic drugs and immune cells to the tumour. As a result, the effectiveness of treatments can be limited. 

Pictured: Conway Fellows, Prof. Brian Rodriguez (left) and Dr Stephen Thorpe (right) with Dr Aidan T. O’Dowling, first author of 'Machine learning and artificial intelligence: enabling the clinical translation of atomic force microscopy-based biomarkers for cancer diagnosis.' (not pictured: Dr Tom Curran, St Vincent's University Hospital).

In addition to this mechanical barrier, the stiffness of the tumour tissue also plays a biological role in promoting tumour progression. Cancer cells within stiffer tissues often respond by activating pro-survival signalling pathways. This helps the cancer cells to continue to grow and survive.

Increased stiffness is linked to more aggressive cancer behaviour, including enhanced invasiveness and a greater likelihood of spreading to other parts of the body.

However, tumour stiffness can vary significantly between patients, and this variation is linked to clinical outcomes. Some patients may have tumours that are particularly stiff and resistant to traditional therapies, while others may not exhibit the same degree of stiffness. The measurement of tumour stiffness has the potential to be an important factor in personalising treatment plans for different patients.

One promising approach for improving cancer treatment is to categorising patients based on the stiffness of their tumours. Clinicians could potentially identify patients who might benefit from therapies that specifically target the tumour's mechanical properties, alongside conventional treatments like chemotherapy and radiotherapy. This combined approach could help overcome the physical and biological barriers posed by stiff tumour tissue.

AFM is a powerful technique that can measure the mechanical properties of tissues at a very fine scale, allowing for the precise assessment of tumour stiffness. The review outlines how AI algorithms could help automate and enhance this process, enabling the clinical adoption of stiffness-based biomarkers for cancer diagnosis and potentially for predicting treatment outcomes.

The research team is particularly focused on the application of these methods to pancreatic cancer, a disease known for its aggressive nature and poor prognosis. They are working to establish the relevance of AFM-based biomarkers for assessing pancreatic cancer tissue stiffness, with the ultimate goal of demonstrating that these biomarkers can be used as reliable diagnostic tools in clinical practice.

By integrating AI and AFM techniques, they hope to improve the ability to predict tumour behaviour and treatment response, leading to more effective and personalised treatment strategies for cancer patients.

Journal citation:
Machine learning and artificial intelligence: Enabling the clinical translation of atomic force microscopy-based biomarkers for cancer diagnosis. 
O’Dowling, Aidan T. et al. Computational and Structural Biotechnology Journal, Volume 24, 661 - 671