IMPLEMENTATION OF A PROGRAMMATIC APPROACH TO SIMPLIFY MASS SPECTROMETRY DATA ANALYSIS BY MACHINE LEARNING AND APPLICATIONS IN BIOMARKER RESEARCH

Abstract

The advancement of biomarker discovery and implementation will enable earlier disease detection, support clinical decision making and improve patient outcomes. A major challenge faced is the development of methods capable of detecting small changes in a biological sample; there is a need for analytical methods that measure changes selectively and sensitively, and data analysis methods that effectively identify those changes. The use of glycans as a biomarker class has unique advantages. Due to their non-template production, their composition is dynamic with changes in the cellular environment. The composition of glycans is extremely heterogenous, so the use of glycans as biomarkers is analytically challenging but could reform the way disease is diagnosed and treated. One way to approach this conundrum is to combine mass spectrometry and machine learning. Mass spectrometry experiments generate great amounts of data, and often times, it is not feasible to peruse in its native size or format. Implementing machine learning with mass spectrometry data allows the inclusion of more data and in turn, will enable advancements in biomarker discovery. A software tool, LevR, has been developed to support mass spectrometrists implementing machine learning into their workflows; this tool provides automated formatting of mass spectrometry data into a machine learning ready format.