Files in this item

FilesDescriptionFormat

application/vnd.openxmlformats-officedocument.presentationml.presentation

application/vnd.openxmlformats-officedocument.presentationml.presentation1407812.pptx (2MB)
PresentationMicrosoft PowerPoint 2007

application/pdf

application/pdf3836.pdf (15kB)
AbstractPDF

Description

Title:MACHINE LEARNING APPLIED TO RAMAN SPECTRA OF PANCREATIC CANCER CELLS TO IDENTIFY MOLECULAR SIGNATURES, PATTERNS AND PROTEIN EXPRESSIONS.
Author(s):Mandrell, Christopher
Contributor(s):Sivakumar, P ; Chowdhury, Farhan ; Abadi, Sakineh E.M.
Subject(s):Spectroscopy as an analytical tool
Abstract:Pancreatic cancer is the third leading cause of cancer-related deaths in the United States, with the life expectancy for patients diagnosed in the late stage ranging from 6-12 months. This study is part of a group effort to examine the hypothesis that, difficult to detect, tumor initiating cells (TIC) exist in small numbers in the solid tumors and are responsible for the cancer’s progression and relapse. The pancreatic cell line MIA PaCa-2 is seeded and grown in 90 Pa 3D fibrin gels for 10 days then samples are collected and Raman spectra obtained from a 784nm laser in the 150-1800nm and 2500-3500nm ranges. These spectra are analyzed via combinations of data preprocessing, wavelength or dimension reduction and machine learning classification algorithms. We extend to other cell lines such as CFPAC-1 and PANC-1. Support vector machine (SVM) and k-nearest neighbors (kNN) supervised machine learning classifiers are applied to the raw and pre-processed data sets and with various statistical and machine learning dimension reduction protocols. These combinations are compared to determine which performed best at classifying cancer and normal cell samples, and which led to selection of the same or similar dimensions. Identification of the best performing dimensions/wavelengths is then attempted from the Raman spectra by comparing them to existing biological molecule Raman databases to identify the patterns in the spectra and any unique molecular signatures or protein expressions that could prove useful to better understanding and therefore treating pancreatic cancer.
Issue Date:2019-06-18
Publisher:International Symposium on Molecular Spectroscopy
Genre:Conference Paper / Presentation
Type:Text
Language:English
URI:http://hdl.handle.net/2142/104291
DOI:10.15278/isms.2019.TL09
Rights Information:Copyright 2019 Christopher Mandrell
Date Available in IDEALS:2019-07-15
2020-01-25


This item appears in the following Collection(s)

Item Statistics