2005 Fiscal Year Final Research Report Summary
Development of Statistical methodology for searching disease related factors based on proteomic and clinical information
| Project/Area Number |
16300090
|
|---|
| Research Category |
Grant-in-Aid for Scientific Research (B)
|
| Allocation Type | Single-year Grants |
|---|
| Section | 一般 |
|---|
| Research Field |
Statistical science
|
|---|
| Research Institution | Japanese Foundation for Cancer Research |
|---|
Principal Investigator |
MATSUURA Masaaki Japanese Foundation for Cancer Research, Cancer Institute, Department of Physics, Associate member, 癌研究所物理部, 主任研究員 (40173794)
|
|---|
| Co-Investigator(Kenkyū-buntansha) |
HOSHIKAWA Yutaka Japanese Foundation for Cancer Research, Genome Center, Gene expression analysis group, Associate, ゲノムセンター, 研究員 (80280626)
MIYATA Satoshi Japanese Foundation for Cancer Research, Genome Center, Bioinformatics Group, Associate, ゲノムセンター, 研究員 (60360343)
USHIJIMA Masaru Japanese Foundation for Cancer Research, Genome Center, Bioinformatics Group, Associate, ゲノムセンター, 研究員 (60328565)
MIKI Yoshio Tokyo Medical and Dental University, Medical Research Institute, Professor, 難治疾患研究所, 教授 (10281594)
EGUCHI Shinto Institute of Statistical Mathmatics, Department of Mathematical Analysis and Statistical Inference, Professor, 統計基礎研究系, 教授 (10168776)
|
|---|
| Project Period (FY) |
2004 – 2005
|
| Keywords | Proteome / Mass Spectrometry / Clinical Information / Statistics / Machine Learning |
|---|
| Research Abstract |
Protein mass spectrometry (MS) is a promising tool for early cancer detection or prediction of anticancer drug sensitivity. An objective of MS is to identify the biomarkers and to enable the prediction of cancer. We have developed a novel methodology consists of three-step strategy for analyses of MS data.
Step (1) Peak detection within individual subjects ; We utilize adaptive free-knot splines with adaptive model selection criterion and local bias/variance estimation procedure. This procedure provides outperforming results compare with previous researches.
Step (2) Peak alignment among subjects. Estimating representative m/z values of common peaks among many subjects is needed. We have developed a simple procedure based on a new idea using information of an adjacent peak in one subject.
Step (3) Statistical analysis for selecting biomarkers and prediction ; We applied boosting alogorithms for cancer classification methods.
We have developed compute programs for these three steps. We reported our study results at American symposium of Advances in Proteomics in Cancer Research in 2004 and Pacific Symposium on Biocomputing at Hawai in 2006. Furthermore, we have had a cooperative symposium at the Institute of Statistical Mathematics and have discussed our issue with Dr.Yasui from Alberta University in Canada.
|
