Poster #37: Suhotro Gorai

GradioMS: a simple tool for machine learning applications on mass spectrometry data analysis

Gorai, S.; University of Delaware

Mass spectrometry has been a crucial technique in protein-based research. It has the ability to provide individual peak signals, which are highly characteristic and help define a protein's individuality, thereby aiding understanding of molecular and cellular biology. This supports modern proteomic studies for identification and quantification with high sensitivity and precision. Modern algorithms using machine learning can help predict and identify characteristic patterns in unknown proteins. However, these methods are primitive and require programming expertise to operate. The lack of simple research tools impedes research for scientists interested in mass spectrometric data analysis. To ease and streamline this process, a Python-based application wrapped using Gradio has been developed. It is readily installable and can be used by researchers without programming expertise to predict protein characteristics using machine learning (ML) methods. This tool houses many interactive modules, where users can first peruse the data, perform baseline correction, perform preprocessing, and run ML models of their choice, including regression, support vector machines, random forests, and so on. Users can save the parameters used and the output generated in the form of graphs for future reference and publications. This tool will aid in performing predictions and extrapolating to any newly discovered or unknown proteins to provide insights into possible functional features.