SHERP

Drug response prediction model using a component based structural equation modeling method
구조방정식을 이용한 약물반응 예측 모형 설립 및 성능 비교

Cited 0 time in webofscience Cited 0 time in scopus
Authors
김성태
Advisor
박태성
Major
자연과학대학 협동과정 생물정보학전공
Issue Date
2017
Publisher
서울대학교 대학원
Keywords
Prediction ModelComponent Based Structural Equation ModelingMultiple Reaction Monitoring (MRM)Generalized Structured Component Analysis(GSCA)
Description
학위논문 (석사)-- 서울대학교 대학원 자연과학대학 협동과정 생물정보학전공, 2017. 8. 박태성.
Abstract
The liver is made up of many different types of cells. Mutations in those cells can be developed into several different forms of tumors as known as cancers. For this reason, it is hard to expect for a single type of liver cancer treatment to have a favorable prognosis for all cancer patients. If we can diagnose and classify the patients who are expected to have good responses to a single therapeutic drug, it will help to reduce the time on choosing appropriate therapeutic drug for each patient health efficiently. Therefore, nowadays, building decent prediction model became important. Up to date, several models such as linear/logistic regression (LR), support vector machine (SVM), random forest (RF) methods have been used for prediction. However, occasionally, these methods oversight the biological pathway information with relations between metabolites, proteins, or DNAs. In this research, we selected possible biomarkers and constructed a drug called Sorafenib response prediction model for liver cancer patients using a component based structural equation model. Component based structural equation modeling method have been used in sciences, business, education and other fields. This method uses unobservable variables as known as latent variables and the structural equation model relationships between variables.
In our research, we applied this structural equation modeling method into biological structured information data. Currently, we have peptide level data with Multiple Reaction Monitoring (MRM) mass spectrometry. MRM is a highly sensitive and selective method for targeted quantitation of peptide abundances in complex biological samples. The advantage of our component based drug response prediction model is that it first merges peptide level data into protein level information which helps better biological interpretation later. Also, it uses alternating least squares algorithm and estimates both coefficients of peptides and proteins efficiently. It handles correlation between variables without constraint by a multiple testing problem. Using estimated peptide and protein coefficients, we have selected significant protein biomarkers with permutation test and constructed a Sorafenib response prediction model. Using drug response for liver cancer patients’ MRM data, we composed a Sorafenib response prediction model for liver cancer and demonstrated that our prediction model successfully predicted a drug response for liver cancer patients with high area under the curve (AUC) score.
Language
English
URI
http://hdl.handle.net/10371/138101
Files in This Item:
Appears in Collections:
College of Natural Sciences (자연과학대학)Program in Bioinformatics (협동과정-생물정보학전공)Theses (Master's Degree_협동과정-생물정보학전공)
  • mendeley

Items in S-Space are protected by copyright, with all rights reserved, unless otherwise indicated.

Browse