Multiscale biochemical mapping of the brain through data-driven and machine learning enabled- mass spectrometry

Xie, Yuxuan

Permalink

https://hdl.handle.net/2142/121309 Copy

Description

Title

Multiscale biochemical mapping of the brain through data-driven and machine learning enabled- mass spectrometry

Author(s)

Xie, Yuxuan

Issue Date

2023-06-28

Director of Research (if dissertation) or Advisor (if thesis)

Sweedler, Jonathan V.

Doctoral Committee Chair(s)

Sweedler, Jonathan V.

Committee Member(s)

• Lam, Fan
• Bhargava, Rohit
• Kraft, Mary L.

Department of Study

Bioengineering

Discipline

Bioengineering

Degree Granting Institution

University of Illinois at Urbana-Champaign

Degree Name

Ph.D.

Degree Level

Dissertation

Keyword(s)

Mass Spectrometry, Machine Learning

Abstract

For brain research, there is an ever-growing demand for new generations of analytical methods that can allow researchers to see beyond the morphology of neural tissues, but to understand the mechanisms of various brain functions and diseases. Mass spectrometry (MS) based measurements both provide unmatched chemical specificity and resolution. Laser-based MS have allowed spatial-omics profiling of tissues through imaging, improved mass resolution for outstanding detection specificity, and achieved attomole detection limits for small volume samples at the single-cell resolution. However, there exist several major technological challenges for comprehensive mapping of the brain biochemistry using MS: extremely low throughput for high- resolution mass spectrometry, lack of quantitative framework to extract biological knowledge, and no current methods to integrate multiscale and multimodal data sets. Here, we present a framework based on computational MS strategies to elucidate multiscale biochemistry of brain at both tissue and single-cell level. We first enabled high-throughput MS imaging via model-based approach to computationally reconstruct high-mass-resolution MS data, significantly accelerating data acquisition and improving signal-to-noise ratio by 10-fold. We further devised a data analysis pipeline based on machine learning to extract biologically relevant information from data collected on individual cells and organelles. Through the pipeline, we were able to classify and interpret the heterogeneity and variability present within and across different brain cell types or organelle types, as well as cells from different brain anatomical regions. Using the computational framework, we performed simultaneous three-dimensional brain-wide and single-cell biochemical mapping containing millions of pixels and large single-cell populations across rat brains. Via multimodal registration and data integration, we created 3D molecular distribution with rich chemical details, and identified cell-specific lipids depending on both cell types and anatomical origins of the cells.

Graduation Semester

2023-08

Type of Resource

Thesis

Handle URL

https://hdl.handle.net/2142/121309

Copyright and License Information

Copyright 2023 Yuxuan Xie

Owning Collections