Introduction to Bioinformatics for Medical Application

Presentation on theme: "Introduction to Bioinformatics for Medical Application"— Presentation transcript:

1 Introduction to Bioinformatics for Medical Application
14 / June / 2007 Hiroyuki Hamada Laboratory for Bioinformatics Hello everybody. My name is Hiroyuki Hamada. I am research assistant of laboratory for bioinformatics. My English is not good enough yet. Please take it easy on me. Well, Prof. Okamoto, he is my boss, Prof. Okamoto is responsible for teaching this lecture. However, he is going to business trip, today. So, I am substitute instructor. If you have any question in this lecture, I would like you to ask Prof. Okamoto the question in his lecture. I hope your kindly cooperation. OK, today, I am talking about Introduction to Bioinformatics for Medical Application. Do you know this term “Bioinformatics”?

2 生物情報学による 医学・工学への アプローチ （我々の研究戦略）
観測データの収集 遺伝子発現量の時系列データ タンパク質合成量の時系列データ 時系列 遺伝子など 測定値 0.0 1.0 表現型（疾患レベルなど） 遺伝子・タンパク質の おおよその機能分類 相互作用の推定 陽性 陰性 gene1 gene2 gene3 gene4 gene5 gene9 gene7 gene11 gene15 判別器 ブーリアンネットワーク ベイジアンネットワーク クラスタリング解析 遺伝的アルゴリズム 遺伝的プログラミング （システム同定） 処方設計 薬剤開発 生物情報学による 医学・工学への アプローチ （我々の研究戦略） 高収率な代謝生産物生産のための 代謝制御系のシステム設計 シミュレーション システムの安定性解析 システムの感度解析 支配因子の同定

3 遺伝子発現データ DNA microarray 実験 遺伝子発現量の経時変化データ Time series graph Time 1
Gene# Expression 0.0 1.0 Time 1 Time 2 Well, in Next session, I would like to show the analysis for biological and medical data with using bioinformatics. Firstly, we have to perform the DNA microaray experiments and collect the time course data of gene expression. Time series of gene expression is illustrated on three dimensional graph. Time 3 Time series graph

4 クラスタリング解析 次元圧縮 標準と標本の比較 階層型クラスタリング クリスプクラスタリング 機能グループの分類 遺伝子間相互作用因子の分類
標準体と変異体の分類 Time Gene# Expression 0.0 1.0 Time Expression 0.0 1.0 This three dimensional graph compress to two dimensional graph with employing clustering analysis. So, we find the functional groups based on the expression pattern. Moreover, if we compare member of each cluster on the samples with member of control, we find the gene interactions and mutations. In the Bioinformatics for medical application, we often adopt whether Hierarchical clustering or k-means clustering. 階層型クラスタリング クリスプクラスタリング

5 クラスタリング法  階層型クラスタリング: ツリー構造 ► Each gene is a leaf on the tree
► Distances reflect similarity of expression ► Internal nodes represent functional groups  クリスプクラスタリング: k-平均クラスタリング ► Number k is chosen in advance ► Each group represents similar expression time Hierarchical clustering generates a tree Each gene means a leaf on the tree Distances reflect similarity of gene expression time. Internal nodes represent functional groups. Therefore Hierarchical clustering sorts the gene with expression time, and forms each cluster with arbitrary threshold as shown in this figure. On the other hand, k-means clustering generates k groups. before starting the analysis, Number k is chosen based on biological findings. Each group represents similar expression. This figure shows the concept of k-means clustering. In the k-means clustering, number of dimension is as same as number of sampling time. If we have two samples dependent on time, two dimensional graph is adopted as shown in this figure. Centroids is optimized as all data include any clusters. Therefore, I think k-means clustering is available for analysis of time course data of gene expression than Hierarchical clustering, Because k-means clustering can taka expressions value into account.

6 クラスタリングの結果  Causes of similar expression between genes are identified as follows: ► One gene controls the other in a pathway ► Both genes are controlled by another ► Both genes related to same time in cell cycle ► Both genes have similar function Then, What does expression correlations mean? Causes of similar expression between genes are identified as follows: One gene controls the other in a pathway. Both genes are controlled by another. Both genes related to same time in cell cycle. Both genes have similar function. Therefore, we should validate and verify the member of each cluster in order to find the functional disorder of gene on the disease. If genes related to disease are identified, we need to clarify the interactions between clusters, perhaps between members in same cluster.

7 システム同定 数理モデルの構築 クラスタ間の相互作用を推定 S-System Genetic programming 標準との比較により
機能不全のクラスタや遺伝子を同定 Time Expression 0.0 1.0 1 2 3 4 5 6 7 8 9 n Clarification of interactions between genes or clusters are called system identification. In system identification, we need to structure the mathematical model. The mathematical model have to represent the time course data of this expression. Mathematical model is design with employing S-system or Genetic programming. Prof Okamoto will explain the S-System and Genetic programming in his lecture, so, today I don’t explain these techniques at detail. S-System and genetic programming are one of the methodologies, and define a mechanism of system with using ordinary differential equation. If we structured the mathematical model, we can infer the interactions between genes or clusters from this time course data. Then, we find the functional disorder clusters or genes by comparing with control. S-System Genetic programming クラスタ間の相互作用を再現する 数理モデルの構築

8 システム解析 支配因子の同定 ボトルネック経路の同定 数値シミュレーション 統計解析 処方構築 創薬 判別器 2 3 4 5 6 7 8 9
　処方構築　 　創薬　 1 2 3 4 5 6 7 8 9 n gene4 gene3 If we structured a mathematical model of system of interest, the mathematical mode can realize the time course data. Then we can perform the computer simulation under the arbitral condition with employing the mathematical model. Everyone knows, computer simulations are available for design of prescription, identification of bottle-neck pathway, discovery of genetic therapy etc. If we find the dominant genes, we can develop the discriminant machine with applying statistical analysis as shown in this figure. These analysis are called system analysis. Clustering analysis, system identification, system analysis are indispensable for advances in medical sciences. These techniques are introduced from bioinformatics in order to solve the problem in medical sciences. Negative gene2 gene15 gene5 gene1 数値シミュレーション 統計解析 gene11 gene5 Positive gene7 gene9 　判別器　

9 まとめ GenBank OMIM BLAST PubMed
Ok, I am talking about Introduction to bioinformatics for medical application without using mathematical techniques. I would like you to understand the concept of bioinformatics for medical application in this lecture. In bioinformatics for medical application, Firstly, we collect the experimentally biological data, next perform the clustering analysis to find the functional groups, Then, identify the mechanism of system, Finaly, try to explore the valuable information on medical sciences with using system analysis. Everywhere, every time, we can integrated these techniques with PubMed, OMIM, GenBank and BLAST. This slide shows the concept of bioinformatics for medical application very well. I wonder if you understand the concept of bioinformatics for medical application.