Trust-MDx - Trustworthy Decision Limits for Multiplexed Diagnostics

I propose a fundamentally I propose a fundamentally new mathematical framework to leverage recent developments in ultrasensitive multiplex diagnostic testing. Early and accurate disease diagnosis has a tremendous influence on patient treatment possibilities, survival rates, and outbreak prevention. It allows for timely and relevant interference by the healthcare system and more effective allocation of medical resources.

In recent years there has been a large amount of effort directed at improving test sensitivity by material scientists. This led to innovative developments utilising various nanoparticles to enhance biomarker signals. The recent use of surface-enhanced Raman spectroscopy has improved the sensitivity of diagnostic tests while allowing for multiplex detection. Unfortunately, the enormous recent developments by material scientists have not been accompanied by similar advances in mathematical modelling, which is why the mathematical framework underpinning the diagnostic tests relies on ineffective century-old modelling techniques.

Common problems in biomarker detection pertain to integrating a single Ramen band unique to the biomarker of interest. This is very costly in terms of sensitivity, which renders the recent developments by material scientists ineffective. Through latent variable-based models, I will increase the sensitivity of multiplex diagnostic tests. This way of modelling will, however, introduce complex sample-specific uncertainty structures, which must be precisely estimated to facilitate useful decision-making based on multiplex diagnostic tests. I mitigate this by utilizing machine-learning principles in a form of gaussian processes to learn the complex uncertainty structures.

This will constitute new mathematical theories and practices to facilitate decision-making based on multiplexed diagnostic tests. I will improve the sensitivity and specificity of multiplexed diagnostic tests, and hence enable precise and early disease diagnosis.