臺灣博碩士論文加值系統

　　對於微陣列實驗與表達序列標籤實驗而言，其目的便是針對研究人員所感興趣的組織細胞中所取出的mRNA作檢測並測量它的表現值，兩者之間所欲達成的目標是相同的，但是mRNA在這兩種實驗中卻有不同的表現量。為了近似並找出微陣列數值與表達序列標籤數值的關係，本系統採用了具有函數近似功能的徑向基底函數類神經網路作為計算兩者關係的方法；而其中包含了四個部份：(1)利用微陣列數值與表達序列標籤數值組成訓練樣本；(2)針對樣本中歪斜的情況進行加權式樣本的產生；(3)利用聚類分析法找出訓練樣本的外圍極端值並移除；(4)最後再利用k-means對樣本空間分析，以找出合適的徑向基底函數中心，然後再進行訓練徑向基底函數類神經網路的動作。當訓練完成的時候，使用者便可以輸入微陣列的表現值，進而由系統計算出所相對的表達序列標籤數值。

　　The　main　purpose　of　both　microarray　and　EST　experiments　are　to　measure　the　mRNA　expression　level　but　in　different　quantity　representation.　The　aim　of　this　study　therefore　is　to　approximate　the　relations　between　two　experimental　measurements.　The　underlying　heuristics　of　our　approach　based　on　RBF　neural　network　with　functional　approximation　mechanism　consisted　of　1)　corresponding　microarray　with　EST　expression　values　to　generate　the　training　samples,　2)　adopting　weighted-pattern　generator　to　resolve　the　skew　distribution　among　the　samples,　3)　identifying　outliers　through　clustering,　and　4)　detecting　centers　of　RBFs　by　analyzing　the　feature　space　of　the　training　samples　using　k-means　algorithm　in　order　to　facilitate　the　training　process　of　neural　network.　Upon　completion　of　neural　network　learning,　system　returns　the　corresponding　EST　count　when　microarray　expression　value　is　submitted.　Our　experiments　have　shown　that　the　overall　error　rate　fell　from　10%　to　30%　when　the　five-dimensional　feature　vector　was　applied.　If　we　are　to　increase　the　neural　network　performance,　more　training　samples　will　be　required.　Furthermore,　the　results　of　feature　selection　comparison　experiments　suggest　that　using　five-dimensional　feature　vector　performed　better　than　that　of　the　three-dimension.　On　the　other　hand,　in　the　outlier　detection　experiments,　less　outliers　were　identified　if　the　size　of　data　set　grew.

第一章　導論..............................................................1
1.1　前言.................................................................1
1.2　研究動機.............................................................2
1.3　解決方法.............................................................3
1.4　系統概述.............................................................4
1.5　論文架構.............................................................5

第二章　文獻回顧與相關研究................................................6
2.1　微陣列製作流程.......................................................6
2.2　表達序列標籤製作流程.................................................9
2.3　微陣列與表達序列標籤數據關聯性之相關研究............................11
2.4　類神經網路於生物領域方面之相關研究..................................12

第三章　應用類神經網路計算微陣列與表達序列標籤數據間的近似關係...........13
3.1　方法論概述..........................................................13
3.2　系統架構圖..........................................................14
3.3　微陣列與表達序列標籤數據資料的前處理................................16
3.3.1　微陣列資料的處理..................................................16
3.3.2　表達序列標籤資料的處理............................................18
3.3.3　訓練樣本的產生....................................................19
3.4　加權式樣本(Weighted-pattern)產生方法................................21
3.5　類神經網路架構方法..................................................26
3.5.1　特徵的擷取........................................................26
3.5.2　神經元高斯函數之中心點選取與權重值更新方法........................29
3.5.3　架構類神經網路....................................................32
3.6　藉由聚類法偵測外圍極端值............................................33

第四章　實驗設計與結果分析...............................................35
4.1　資料集介紹及測試方法介紹............................................35
4.2　實驗與結果分析......................................................36
4.2.1　類神經網路與非線性?歸比較.........................................36
4.2.2　特徵選取之比較....................................................40
4.2.3　偵測外圍極端值策略的比較..........................................45

第五章　結論與未來研究方向...............................................49
5.1　結論................................................................49
5.2　未來研究方向........................................................49

參考文獻.................................................................51

[1]　A.　Coghlan.,　et　al.,　“Relationship　of　codon　bias　to　mRNA　concentration　and　protein　length　in　Saccharomyces　Cerevisiae,”　Yeast,　16:1131-1145,　2000.
[2]　A.　D.　Baxevanis,　et　al.,　“A　practical　guide　to　the　analysis　of　genes　and　proteins,”　Second　edition.
[3]　A.　Eyre-Walker,　“Synonymous　codon　bias　is　related　to　gene　length　in　Escherichia　Coli:　selection　for　translational　accuracy,”　Molecular　Biology　Evolution,　13:864-872,　1996.
[4]　A.　Papoulis,　“Probablity,　random　vairables　and　stochastic　Process,”　Fourth　edition.
[5]　C.　Li,　W.　H.　Wong,　“Model-based　analysis　of　oligonucleotide　arrays:　expression　index　computation　and　outlier　detection,”　PNAS,　98:31-36,　2001.
[6]　D.　Yoon,　et　al.,　“Two-stage　normalization　using　background　intensities　in　cDNA　microarray　data,”　BMC　Bioinformatics,　5:97,　2004.
[7]　E.　T.　Munoz,　et　al.,　“Microarray　and　EST　database　estimates　of　mRNA　expression　levels　differ:　the　protein　length　versus　expression　curve　for　C.　elegans,”　BMC　Genomics,　5:30,　2004.
[8]　J.　B.　Tobler,　et　al.,　“Evaluating　machine　learning　approaches　for　aiding　probe　selection　for　gene-expression　arrays,”　Bioinformatics,　March　27,　2002.
[9]　J.　H.　Kim,　et　al.,　“Effect　of　local　background　intensities　in　the　normalization　of　cDNA　microarray　data　with　a　skewed　expression　profiles,”　Experimental　and　Molecular　Medicine,　Vol.　34,　No.　3,　224-232,　July　2002.
[10]　J.　M.　Deutsch,　“Evolutionary　algorithms　for　finding　optimal　gene　sets　in　microarray　prediction,”　Bioinformatics,　July　12,　2002.
[11]　J.　P.　Townsend,　“Multifactorial　experimental　design　and　the　transitivity　of　ratios　with　spotted　DNA　microarrays,”　BMC　Genomics,　4:41,　2003.
[12]　L.　Duret,　et　al.,　“Statistical　analysis　of　vertebrate　sequences　reveals　that　long　genes　are　scarce　in　GC-rich　isochors,”　Journal　of　Molecular　Evolution.,　40,　308,　V317.　1995.
[13]　M.　S.　B.　Sehgal,　“Collateral　missing　value　imputation:　a　new　robust　missing　value　estimation　algorithm　for　microarray　data,”　Bioinformatics,　February　24,　2005.
[14]　M.　Semon,　et　al.,　“Relationship　between　gene　expression　and　GC-content　in　mammals:　statistical　significance　and　biological　relevance,”　BMC　Genomics,　4:49,　2003.
[15]　Protech　Technology,　“Gene　discovery　introduction,”　ProNews　23,　Jul,　2004.
[16]　R.　Linder,　et　al.,　“The　subsequent　artificial　neural　network(SANN)　approach　might　bring　more　classificatory　power　to　ANN-based　DNA　microarray　analyses,”　Bioinformatics,　July　29,　2004.
[17]　R.　Soumya,　et　al.,　“Principal　components　analysis　to　summarize　microarray　experiments:　application　to　sporulation　time　series,”　Stanford　Medical　Informatics,　2000.
[18]　S.　Alexander,　et　al.,　“A　comprehensive　evaluation　of　multicategory　classification　methods　for　microarray　gene　expression　cancer　diagnosis.,”　Bioinformatcs,　September　16,　2004.
[19]　S.　Haykin,　“Neural　networks:　a　comprehensive　foundation,”　Sencond　edition.
[20]　T.　C.　Santiago,　et　al.,　“The　relationship　between　mRNA　stability　and　length　in　Saccharomyces　Cerevisiae,”　Nucleic　Acids　Res.,　14:8347-8360,　1986.
[21]　W.　Zhang,　et　al.,　“The　functional　landscape　of　mouse　gene　expression,”　BMC　Journal　of　Biology,　Dec　6,　2004.
[22]　Cancer　Genome　Anatomy　Project,　CGAP,　http://cgap.nci.nih.gov/
[23]　dbEST,　http://www.ncbi.nlm.nih.gov/dbEST/
[24]　Introduction　to　Biotechnology,　http://juang.bst.ntu.edu.tw/JRH/biotech.htm
[25]　Mammalian　Gene　Collection,　MGC,　http://mgc.nci.nih.gov/
[26]　National　Center　for　Biotechnology　Information,　NCBI,　http://www.ncbi.nlm.nih.gov/
[27]　Nucleotide,　http://www.ncbi.nlm.nih.gov/entrez/query.fcgi?db=nucleotide
[28]　Stanford　Microarray　Database,　SMD,　http://genome-www5.stanford.edu/
[29]　UniGene　Digital　Differential　Display,　DDD,　http://www.ncbi.nlm.nih.gov/UniGene/info_ddd.html
[30]　教育部生物科技教育改進計畫，http://abep.sinica.edu.tw/stretagy7.htm
[31]　莊順淑，生物資訊概論，http://www.ascc.net/nl/92/1920/02.txt
[32]　陳健尉，生物晶片教學，http://www.nchu.edu.tw/~ibms/jwlaboratory.htm