INFORM - non-invasive early detection of brain cancer using tear fluid proteome profiling (supported by Stand Up To Cancer)

Background: Despite being considered a rare disease, brain cancer represents a significant socio-economic burden, with high mortality rates and disabling effects on patients.

Since the risk of recurrence and mortality are closely related to the disease stage at the time of primary intervention, late diagnosis is among the key factors contributing to the poor outcome of brain cancer patients.

Therefore, there is a significant unmet need to improve patient care by introducing a diagnostic ‘triaging’ tool that is accessible in a timely fashion.

Our research team found a correlation between intracranial tumour growth, the dynamics of immune cells residing in the ocular space and changes in tear fluid proteome. We hypothesized that tear-fluid biomarkers can be used for a non-invasive and early detection of brain cancer.

A prospective pilot study (n=133 human subjects including gliomas, brain metastases and healthy volunteers) carried out by our team led to the development of a machine-learning model which detected brain cancer with 96% weighted accuracy, 100% sensitivity and 91.7% specificity.

Aims: We aim to develop a tear protein-based machine learning model that can detect and diagnose brain cancer at an early stage with a high degree of sensitivity and specificity.

Methods: We will combine unbiased discovery (mass spectrometry & data-independent acquisition) with targeted (OLINK) proteomics followed by machine learning (ML) to accelerate tear-fluid biomarker translation and the development of a tear protein-based ML model.

We will test the analytical performance of this ML model on an unseen blinded validation cohort and assess its readiness for deployment using standard metrics such as accuracy, precision-recall, area under the ROC curve, F1 score.

How the results of this research will be used: If successful this project will deliver a tool that can help physicians to diagnose brain cancer early with high accuracy and speed, without the need for invasive biopsy procedures or costly MRI.

In addition, this tool is expected to assist with classifying brain cancer into different types and monitor their progression and response to various treatments, such as surgery, radiation, and chemotherapy.