【論】Hanisch,2002,Co-clustering of biological net～

Daniel Hanisch, Alexander Zien, Ralf Zimmer and Thomas Lengauer
Co-clustering of biological networks and gene expression data
Bioinformatics Vol. 18 no. 90001 2002, Pages S145-S154
[PDF][Web Site]

・遺伝子ネットワーク推定法について。従来法は遺伝子発現量情報のみに基いていたのに対して、さらにBiological networksの情報を付加した方法を提案する。
・データ：Yeast, 時間点7点 [DeRisi,1997] Metabolic networkに着目。
・クラスタリング法：Hierarchical average linkage clustering [Eisen,1998]。閾値の決定はSilhouette-coefficient [Rousseeuw,1987]による。

・問題点「One major problem is that the boundaries of the resulting clusters are arbitrary to some degree. 」
・方法「In this paper, we propose a novel method that utilizes information in the form of biological networks in an integrated manner to improve the result of the clustering.」
・「Biological networks relate genes, gene products or groups of those (e.g., protein complexes or protein families) to each other in the form of a graph (Bhalla and Iyengar,1999).」
・方法「Whereas conventional cluster methods rely on gene expression data alone, we propose combining measures derived from gene expression data and metrics on biological networks into a single distance function.」
・「The most popular choice in the case of time-series data is the Pearson correlation coefficient, as suggested by Eisen et al.(1998).」

【論】He,2006,In search of functional association ～

Feng He and An-Ping Zeng
In search of functional association from time-series microarray data based on the change trend and level of gene expression
BMC Bioinformatics. 2006; 7: 69.
[PDF]

・マイクロアレイ時系列データより遺伝子ネットワークを推定する方法を提案する。各遺伝子の発現量の時間変化量同士の相関を計算し、相関の高い遺伝子のペアを抽出する。
・比較した方法
1. Trend correlation (TC) method (提案法)
2. Local clustering (LC) method
3. Pearson correlation coefficient (PCC) based clustering method
・データ：Yeast cell cycle, over 6000 ORFs, 17 time points [Cho]

・方法「The principle of our method is to use information in the change trend and the change level of gene expression between consecutive time points for the inference of functional linkages among genes.」

・従来法はデータを点で見ていたところを、提案法では連続（trend）として見るとのことですが、この"trend"の考え方がいまいちピンとこない。
・図[写真]中の遺伝子（グラフ）同士には関連があるとのことですが、パッと見る限りでは・・・怪しげ・・・ホンマカイナ。

【論】Stephanopoulos,2002,Mapping physiological sta～

Gregory Stephanopoulos, Daehee Hwang, William A. Schmitt, Jatin Misra and George Stephanopoulos
Mapping physiological states from microarray expression measurements
Bioinformatics Vol.18 no.8 2002 Pages 1054-1063
[PDF][Web Site]

・マイクロアレイデータから、Fisher discriminant analysisに基いて各サンプルのphysiological statesを判定し、図示化する方法を提案する。
・データ
1. normal/malignant oral epithelium [Alevizos,2001]
2. ALL/AML [Golub,1999]
3. cultures of the photosynthetic bacterium Synechocysis sp.PCC 6803 [Gill,2002]
4, single gene deletion mutants of yeast [Hughes,2000]

・方法「The method uses Fisher discriminant analysis to create a linear projection of gene expression measurements that maximizes the separation of different sample classes.」
・FDAとは「FDA is a classification method that operates by determining a set of orthonormal dimensions where the separation between the given classes is maximized.」

【論】Konig,2006,Discovering functional gene express～

Rainer König, Gunnar Schramm, Marcus Oswald, Hanna Seitz, Sebastian Sager, Marc Zapatka, Gerhard Reinelt and Roland Eils
Discovering functional gene expression patterns in the metabolic network of Escherichia coli with wavelets transforms
BMC Bioinformatics 2006, 7:119
[PDF][Web Site]

・マイクロアレイデータより遺伝子間のネットワークを推定する。具体的には、E.coliのaerobic/anaerobic環境下のデータより、metabolicのネットワークを推定する。

・処理「We mapped gene expression data from E. coli under aerobic and anaerobic conditions onto the enzymatic reaction nodes of its metabolic network.」
・マイクロアレイ研究のまとめ「However, the advent of DNA microarrays has allowed us to explore a major subset or all genes of an organism under a variety of conditions such as alternative treatments, mutants, developmental stages and time points. For example, the technique enables us to classify tumor samples [5], to define small sets of potential marker genes to distinguish leukemias [6], and to discover regulatory mechanisms [7,8]. E.g., without prior information, the structure and function of the network that regulates the SOS pathway in E.coli could be elucidated with transcription profiles [9]. Furthermore, physical and chemical interaction data of proteins have been integrated. Knowledge of protein-protein interaction from high-throughput techniques [10] was applied to analyse gene expression data and revealed novel regulatory circuits [11]. Moreover, interaction knowledge from the biochemical network has been used to support the clustering procedure for gene expression profiles of yeast [12,13].」
・まとめ「Hence, we elucidated some interesting and relevant subgraphs of the metabolic network that showed necessary changes during the aerobic - anaerobic shift. But note, that such findings may not represent the entire regulatory change during such a shift of the metabolic network.」
・「Such a "Haar" wavelet transform can be regarded as a low pass filter when calculating the mean, and a high pass filter when calculating the difference between neighbouring value pairs.」

・遺伝子ネットワークの図は、生物学の知識がないのでさっぱり分からない。
・ウェーブレット変換に関する記述はほとんどなし。

【論】Long,2001,Improved statistical inference from ～

Anthony D.Long, Harry J.Mangalam, Bob Y.P.Chan, Lorenzo Tolleri, G.Wesley Hatfield, and Pierre Baldi
Improved Statistical Inference from DNA Microarray Data Using Analysis of Variance and A Bayesian Statistical Framework
J. Biol. Chem., Vol. 276, Issue 23, 19937-19944, June 8, 2001
[PDF][Web Site]

・ベイズ統計を用いた遺伝子抽出法について。またそれを応用して開発したマイクロアレイデータ解析ソフト、Cyber-T（http://visitor.ics.uci.edu/genex/cybert/）について。
・データ：IHF(integration host factor)+ and IHF- Escherichia coli cells[Arfin,2000]
・比較した統計法：Bayesian t test, t test, ratio of means, difference of means

・概要「We also show that statistical tests based on analysis of variance and a Bayesian prior identify genes that are up- or down-regulated following an experimental manipulation more reliably than approaches based only on a t test or fold change.」
・問題点「This creates problems of statistical inferences because many genes will show fairly large changes in gene expression purely by chance alone.」
・問題点「Commonly used software packages are poorly suited for implementing the Bayesian statistical methods we develop in this work.」
・注意「Once the data are log-transformed the standard deviation in gene expression is larger for lowly expressed genes than for highly expressed genes. This relationship makes biological sense because it is difficult to accurately measure and quantify genes showing very low expression levels using high density arrays.」
・注意「Although the consistency of different statistical approaches appears fair with modest levels of replication, it should be noted that the different statistical approaches are not necessarily consistently identifying the same set of genes.」
・結果「Interestingly, in these plots it appears that the variance in log-transformed expression levels is higher for genes expressed at low or near background levels may be good candidates for ignoring in expression analyses.」

【論】Li,2000,Model-based analysis of oligonucleotid～

Cheng Li and Wing Hung Wong
Model-based analysis of oligonucleotide arrays: Expression index computation and outlier detection
Proc.Natl.Acad.Sci. Vol.98, 31-36. 2001
[PDF][Web Site]

・Model-based解析について。ノイズを含んだマイクロアレイデータから、統計的手法で本来の発現量を推定する。
・データ：HuGeneFL array × 21個。

・概要「Here, we propose a statistical model for the probe-level data, and develop model-based estimates for gene expression indexes.」
・問題点「Besides the original publications by Affymetrix scientists, there have been very few studies on important "low-level" analysis issues such as feature extraction, normalization, and computation of expression indexes.」
・問題点「In addition, human inspection and manual masking of image artifacts is currently very time consuming and represents a limiting factor in large-scale expression profiling projects.」
・処理「These include acconting for individual probe-specific effects, and automatic detection and handling of outliers and image artifacts.」

【論】Irizarry,2003,Exploration, normalization, and ～

Rafael A.Irizarry, Bridget Hobbs, Francois Collin, Yasmin D. Beazer-Barclay, Kristen J.Antonellis, Uwe Scherf and Terence P.Speed
Exploration, normalization, and summaries of high density oligonucleotide array probe level data
Biostatistics 4:249-264 (2003)
[PDF][Web Site]

・GeneChipの包括的調査（RMAの性能を検証？）。
・データ
1．a small experimental study consisting of five MGU74A mouse GeneChip arrays
2. part of the data from an extensive spike-in study conducted by Gene Logic and Wyeth's Genetics Institute involving 95 HG-U95A human GeneChip arrays
3. part of a dilution study conducted by Gene Logic involving 75 HG-U95A GeneChip arrays
・発現量指標（summary measures）
1. average difference(AvDiff)
2. MAS 5.0
3. Li and Wong multiplicative model-based expression index (MBEI)
4. robust multi-array average (RMA)

・概要「In this paper we report exploratory analyses of high-density oligonucleotide array data from the Affymetrix GeneChip? system with the objective of improving upon cyrrently used measures of gene expression.」
・RMA「To summarize, in this paper we consider a new expression measure that (i) background-corrects the arrays using the transformation B(・), (ii) normalization the arrays using quantile normalization , and (iii) for each probe set n, fits a linear model (2) to the background-corrected, normalized and log (base 2) transformed probe intensities」
・「The advantage of RMA is especially noticeable in the low expression values where the SD is 10 times smaller than the other measures.」
・結論「We conclude that there is no obvious downside to summarizing the expression level of a probe set with RMA, and attaching an SE to this quantity using a linear model that removes probe-specific affinities.」

【論】Zien,2001,Centralization: a new method for the～

Alexander Zien , Thomas Aigner , Ralf Zimmer and Thomas Lengauer
Centralization: a new method for the normalization of gene expression data
Bioinformatics Vol.17 no.90001 (2001) Pages S323-S331
[PDF][Web Site]

・新たに開発した遺伝子発現量データの正規化法である、Centralizationの紹介。
・比較対象：Total RNA, Housekeeping, Globalozation, ANOVA

・Centralizationとは「Centralization is a new two-step method for the computation of such normalization factors that is both biologically better motivated and more robust than standard approaches. First, for each pair of arrays the quotient of the constants of proportionality is estimated. Second, from the resulting matrix of pairwise quotients an optimally consistent scaling of the samples is computed.」
・概要「Since a gold standard for the evaluation of normalization is unavailable, we demonstrated the robustness of centralization and the weaknesses of common competing methods on a real life gene expression data set.」

【論】Baldi,2001,A Bayesian framework for the analys～

Pierre Baldi and Anthony D.Long
A Bayesian framework for the analysis of microarray expression data: regularized t-test and statistical inferences of gene changes
Bioinformatics Vol.17 no.6 2001 Pages 509-519
[PDF][Web Site]

・ベイズ統計のマイクロアレイ解析への応用。
・1.Bayesian approach 2.simple fold approach 3.straight t-test　を比較。
・紹介されている処理はR用のソフトウェア、Cyber-Tで利用可能（http://visitor.ics.uci.edu/genex/cybert/）。

・問題点「Current methods are unsatisfactory due to the lack of a systematic framework that can accommodate noise, variability, and low replication often typical of microarray data.」
・目的「Our goal here is to develop a general Bayesian statistical framework for the analysis of array data.」
・アレイデータ解析の三段階「Gene expression array data can be analyzed on at least three levels of increasing complexity. First, the level of single genes, where one seeks to establish whether each gene in isolation behaves differently in a control versus a treatment situation. The second level considers gene combinations, where clusters of genes are analyzed in terms of common functionalities, interactions, co-regulation, and so forth. The third level attempts to infer the underlying regulatory regions and gene/protein networks that ultimately are responsible for the patterns observed.」
・ベイズの定理「Bayes theorem: P(M|D) = P(D|M)P(M)/P(D), where P(D|M) is the data likelihood and P(M) is the prior probability capturing any background information one may have.」

・"ベイズ統計"の何たるかがわかってないので、全く歯が立たない。

【論】Fushimi,2002,Measurement of violin plate vibra～

Akira Fushimi, Keinosuke Nagai and Koichi Mizutani
Measurement of violin plate vibration by acoustical holography using boundary element method
Acoustical Science and Technology Vol.23(2002), No.5 pp.258-266
[PDF][Web Site]

・ヴァイオリンの表・裏板の振動の様子を測定し、視覚化する。
・コンデンサマイクで測定した音圧（sound pressure, 従来法）とPCHHS（提案法）で測定したparticle velocity[写真]で視覚化した結果を比較する。
・楽器を鳴らすのは弓で弾くのではなく、Vibratorで特定の周波数の振動を与える方法で行う。

・概要「We apply the pressure-based conformal holography with a hologram and a source surface coupling (PCHHS) as BEM-based acoustical holography in order to measure the vibration eigenmodes of the actual arching violin plates. We reconstruct and visualize the distributions of the sound pressure and the particle velocity on the surface of the violin plate from the measured hologram data in the sound field.」
・問題点「the application to the general curved vibrating surface such as a musical instrument is quite difficult.」

・たまには気分を変えて、全く異分野の論文を。たまたま見つけたヴァイオリンに関する論文。
・与える周波数（eigenmodes）によって、板の振動の仕方がかなり変わってくるのはわかるのだけど、それと楽器から出る音（印象）とどういった関係があるのかについての言及はナシ。複雑な形状の振動体として格好のモデルではありますが、その辺の言及が無い限りは楽器を使う必然性が薄くなってしまう。
・eigenmodesの概念がイマイチよくわからない（→参考文献[1]）。
・使用の楽器が中国製の安物かイタリア製の高級品かの記述ナシ。両者の比較なんていうのも興味あるところです。

[1]C.M.Hutchins,"The acoustics of violin plates," Sci.Am.,245,126-135(1981).