Machine Learning Models for Predicting Tumor and Overexpressed HSP70 Protein Brain Cells using Raman Spectra

In this paper, we propose methods for preprocessing and analyzing the Raman spectra of the brain using the example of a human brain affected by glioma and a mouse brain subjected to overexpression of a heat shock protein weighing 70 kDa HSP70. Machine learning models were selected to analyze and classify the spectra of healthy and tumor brain cells. Based on the obtained models, maps of precancerous brain regions with prediction of the presence and localization of tumor cells were built. Due to the high penetrating ability of tumor cells into healthy tissues, tumor boundaries are of great interest for research and development of an optimal method of tissue differentiation. Obtaining a model predicting the localization of tumor cells in the near-tumor region of the brain opens up prospects for creating intraoperative equipment to help surgeons perform operations. A method has been proposed for analyzing differences in the spectra of the cortex and striatum with overexpression of HSP70 into the interstitial space and into the cytoplasm, with the help of which it is possible to obtain a rapid and automated description of differences in chemical composition. A classification of brain spectra with HSP70 overexpression was also selected. The HSP70 protein is currently being actively researched worldwide as a possible cure for neurodegenerative diseases such as Alzheimer's disease, Parkinson's disease, Huntington's disease and so on. Studying the raman scattering spectra of the cortex and striatum with different HSP70 overexpression will provide an understanding of the mechanism of heat shock proteins and their effects on the body.