BIOINFORMATICS SEMINAR SERIES
https://bioinformatics.udel.edu/seminar
Dictionary-Guided Transformer for Mass Spectral Classification of Airborne Pathogens
Kyle Regan
PhD candidate, Bioinformatics Data Science
CBCB
University of Delaware
Abstract:
Matrix Assisted Laser Desorption & Ionization (MALDI) Time-of-Flight (ToF) Mass Spectrometry is a powerful analytical tool for identification of microbes through biomarkers such as proteins, peptides, and lipids. MALDI is a soft-ionization technique allowing for the ability to analyze large biomolecules without fragmentation. Mass spectrometry measurements typically require mixing the analyte sample with a matrix onto a metal plate and into a mass spectrometer. The measurement process and size of the mass spectrometers make it necessary for measurements to be performed in an isolated laboratory environment. Zeteo Tech Inc., is developing a portable mass spectrometer capable of obtaining single particle samples from the air for real-time detection of airborne microbes. Real-time microbial detection enables the identification of pathogens in the environment to safe-guard our community from infectious diseases.
The portable mass spectrometer randomly samples atmospheric particles, which are predominantly harmless background particles (i.e. dust, fog, water vapor) with few particles corresponding to pathogenic microbes. Each particle generates a noisy mass spectrum that has to be identified independently posing a significant challenge even for deep learning models. Denoising methods developed for mass spectrometry in controlled laboratory settings fail to translate to environmental settings. However, mass spectra can be represented in a sparse domain using the Sparse Synthesis model, where a carefully constructed dictionary provides side-information to a deep learning model to enhance classification accuracy.
Bio:
Kyle Regan is a PhD Candidate in the Bioinformatics Data Science program at the Center for Bioinformatics and Computational Biology at the University of Delaware. He works with Dr. Gonzalo Arce in the Computational Imaging and Machine Learning Laboratory focusing on implementing machine learning techniques to biodefense and biomedical applications. He graduated from the University of Delaware in 2023 with an Honors Bachelor of Electrical Engineering and a minor in Computer Science. His graduate research focuses on the detection of airborne microbes and proteins from single particle aerosol mass spectrometry using machine learning and signal processing.
