표준화 후 유의한 유전자 선택 방법 조합을 이용한 마이크로어레이 분류 시스템 구현 = (A)implementation of system on microarray classification using combination of significant gene selection method after normalization

다국어 초록 (Multilingual Abstract)

Microarray technology provides large-scale gene expression profiles according to experimental conditions such as quantitative values. The transcription of genes are triggered or inhibited by complex biological interactions and associations. To understand not only the function and the role of a gene in a cell but also the mechanism of cellular phenomena, we need to capture snapshots of cellular process using microarray technology. Nowadays, microarray technology is expected to contribute to filling the ultimate purpose of the total study in bioinformatics. Microarray technology allows the monitoring of expression levels for thousands of genes simultaneously. This novel technology helps us to understand gene regulation as well as a gene by gene interactions more systematically.
In the microarray experiment, however, many undesirable systematic variations are observed. Normalization is the process of removing some sources of variation which affect the measured gene expression levels. Normalization plays an important role in the earlier stage of microarray data analysis. The subsequent analysis results are highly dependent on normalization.
Microarray, based on its gene expression data information, has been applicable to the field of cancer diagnosis with the computer-aided classification and prediction technology. Current clinical practice involves an experienced hematopathologist's interpretation of the tumor's morphology. In this way, however, classification remains imperfect and errors do occur. So, it has been suggested that such microarrays could provide a tool for correct cancer classification. But from a datamining-based point of view, there is one difficulty in microarray data analysis that the number of samples is very small while the number of attributes(i.e., genes) is very large. Therefore, to classify the subtypes of cancer correctly using current microarray technology, we should select the informative genes whose expression pattern was strongly correlated with the class distinction to be predicted.
Independently separated informative genes can contribute to inspiring the study of medical cure after the correct classification by them. These informative genes list data suggest that genes useful for cancer class prediction may also provide insight into cancer pathogenesis and pharmacology. So for this application, we have to separate this informative genes list independently. And that list should be consistent, trusted, and strongly correlated with the class distinction to be predicted.
In this paper, the system that can create the separated informative genes list after Lowess normalization is proposed. The effectiveness of these system and method was evaluated through some experiments. In the experimental results using MLP(Multi-Layer Perceptron)-based classifier system, it was found that the proposed system and the suggested combination method create an independently separated informative genes list with consistency, trust, and strong correlation with the class distinction for microarray data.
Therefore, the proposed system and the suggested combination method in this paper are expected to contribute to providing insight into cancer pathogenesis and pharmacology as well as correct classification of cancer types.

목차 (Table of Contents)

• 표목차 = ⅳ
• 그림목차 = ⅴ
• ABSTRACT = ⅶ
• Ⅰ. 서론 = 1
• Ⅱ. 관련연구 = 4
• A. 마이크로어레이 개요 = 4
• B. 마이크로어레이의 종류 = 9
• 1. cDNA 마이크로어레이 = 10
• 2. 올리고뉴클레오티드 마이크로어레이 = 12
• C. 마이크로어레이의 제작 방법 = 15
• 1. 애피메트릭스 올리고뉴클레오타이드 마이크로어레이 = 15
• 2. cDNA 마이크로어레이 = 19
• D. 마이크로어레이 데이터의 특징 = 22
• E. 마이크로어레이 기반의 기존 암 분류 연구 = 24
• 1. 현대 암 진단 방법 = 24
• 2. 분자 생물학을 이용한 암 분류 = 24
• 3. 마이크로어레이를 이용한 암 분류 연구 분야 = 25
• 4. 요구되는 최소 시스템의 개요 = 26
• 5. 마이크로어레이 기반 암 분류 연구 사례 = 27
• a. Golub의 방법 = 27
• b. 다중 분류기 시스템을 이용한 방법 = 30
• Ⅲ. 유의한 유전자 선택 방법 = 34
• A. 마이크로어레이의 표준화 = 34
• 1. 포괄적인 표준화 방법 = 38
• 2. 특성화한 표준화 방법 = 40
• B. 유의한 유전자 선택 = 45
• C. 조합 방법 = 50
• Ⅳ. 신경망 = 52
• A. 신경망 = 52
• B. 지도 학습과 비지도 학습 = 56
• 1. 지도 학습 = 56
• 2. 비지도 학습 = 57
• C. 신경망 학습 방법 = 59
• 1. 퍼셉트론 = 59
• a. 단순 퍼셉트론 = 60
• b. 다층 퍼셉트론 = 61
• Ⅴ. 실험 = 63
• A. 제안하는 시스템 구조도 = 63
• B. 실험 데이터 = 65
• C. 실험 결과 및 고찰 = 66
• 1. 상위 200개 유전자에 대한 다양한 조합에 따른 분류 성능 = 68
• 2. 상위 90개 유전자에 대한 다양한 조합에 따른 분류 성능 = 71
• 3. 상위 50개 유전자에 대한 다양한 조합에 따른 분류 성능 = 75
• 4. 평균 실행 시간 = 79
• Ⅵ. 결론 및 향후과제 = 80
• 각 슬라이드에서 잘못 분류된 유전자 정보 = 82
• 참고문헌 = 86