A Method for Analyzing Time Course Multi-factor Expression Data with Applications to A Burn Study Baiyu Zhou Department of Statistics Stanford University.

Slides:



Advertisements
Similar presentations
Yinyin Yuan and Chang-Tsun Li Computer Science Department
Advertisements

L Exercise Training and Body Composition Training.
Darya Chudova, Alexander Ihler, Kevin K. Lin, Bogi Andersen and Padhraic Smyth BIOINFORMATICS Gene expression Vol. 25 no , pages
Integrating Cross-Platform Microarray Data by Second-order Analysis: Functional Annotation and Network Reconstruction Ming-Chih Kao, PhD University of.
Metabolism. Feasting Feasting adds to body stores of carbohydrate and fat Excess carbohydrate  used to fill glycogen stores  excess glucose stored as.
. Inferring Subnetworks from Perturbed Expression Profiles D. Pe’er A. Regev G. Elidan N. Friedman.
1 Cell-Cell Interactions Chapter 7. 2 Outline Cell Signaling Receptor Proteins – Intracellular Receptors – Cell Surface Receptors Initiating the Intracellular.
Part IV Cellular Communication Cellular signal transduction The cell cycle and programmed cell death Germ cells and fertilization Cell interactions in.
The Inheritance of Complex Traits
Cancer Biology. 2 Outline 1.How do cancer cells differ from normal cells? Tumor progression Molecular basis for cancer.
4.A.3 Cell Specialization Interactions between external stimuli and regulated gene expression result in specialization of cells, tissues and organs.
Principles of Immunology T Cell Development 3/14/06 “For every problem there is a neat, simple solution, and it is always wrong. “ H L Mencken.
Review Lecture II. 3 pathways to degradation in the lysosome.
‘Gene Shaving’ as a method for identifying distinct sets of genes with similar expression patterns Tim Randolph & Garth Tan Presentation for Stat 593E.
. Differentially Expressed Genes, Class Discovery & Classification.
Multidimensional Analysis If you are comparing more than two conditions (for example 10 types of cancer) or if you are looking at a time series (cell cycle.
Data analytical issues with high-density oligonucleotide arrays A model for gene expression analysis and data quality assessment.
Cell and Molecular Biology Fifth Edition CHAPTER 7 Part 3 Interactions Between Cells and Their Environment Copyright © 2008 by John Wiley & Sons, Inc.
The Microenvironment, Stem Cells, and Cancer. Microenvironment Signaling molecules – G-CSF – Erythropoietin Cell-cell contact – Adherens junctions – Gap.
Computational Analysis of USA Swimming Data Junfu Xu School of Computer Engineering and Science, Shanghai University.
Formation of Regulatory Patterns During Signal Propagation in a Mammalian Cellular Network ROBERT D. BLITZER, RAVI IYENGAR.
LEARNING PROGRAMME Hypothesis testing Intermediate Training in Quantitative Analysis Bangkok November 2007.
Getting the story – biological model based on microarray data Once the differentially expressed genes are identified (sometimes hundreds of them), we need.
Alignment and classification of time series gene expression in clinical studies Tien-ho Lin, Naftali Kaminski and Ziv Bar-Joseph.
Learning Target: Human Body Organization I Can…Explain how the levels of human body organization are created and organized to construct a precise functioning.
Biostatistics in Practice Peter D. Christenson Biostatistician LABioMed.org /Biostat Session 6: Case Study.
Characteristics of Cancer. Promotion (reversible) Initiation (irreversible) malignant metastases More mutations Progression (irreversible)
Expression Modules Brian S. Yandell (with slides from Steve Horvath, UCLA, and Mark Keller, UW-Madison)
Weighted models for insulin Detected by scanone Detected by Ping’s multiQTL model tissue# transcripts Islet1984 Adipose605 Liver485 Gastroc404 # transcripts.
©Edited by Mingrui Zhang, CS Department, Winona State University, 2008 Identifying Lung Cancer Risks.
Huidi Liu, M.D. & Ph.D Genomics Research Centre Harbin Medical University, China Reduced expression of SOX7 in ovarian cancer: a novel.
 Blog questions from last week  hhdstjoeys.weebly.com  Quick role play on stages of adulthood  Early Middle Late  Which component of development are.
吳 華 席 Hua-Hsi Wu, MD OB/GYN, VGH-TPE Aug 12, 2008
S factor  body weight during WM  body weight during ER Supplementary figure 1: S factor distribution in each diet. S.
Passive vs. active transport Passive transport is simply transport down an electrochemical gradient until equilibrium is reached Active transport results.
Jan 18, 2008 Ju Han, Hang Chang, Mary Helen Barcellos-Hoff, and Bahram Parvin Lawrence Berkeley National Laboratory Multivariate.
Chapter 16 Nutrition and Aging: The Adult Years
Concept 18.4: A program of differential gene expression leads to the different cell types in a multicellular organism.
KEY CONCEPT The human body has five levels of organization.
Tissue dynamic and Morphogenesis Dept. Physiology Chang Gung University J. K. Chen, Professor.
Statistics for Differential Expression Naomi Altman Oct. 06.
N318b Winter 2002 Nursing Statistics Specific statistical tests Chi-square (  2 ) Lecture 7.
Te Pas et al Additional File 3 Detailed graphical representation of the main pathway results.
Hierarchy of organization Cells → Tissues → Organs → Organ Systems → Organism.
Personal Control over Development: Effects on the Perception and Emotional Evaluation of Personal Development in Adulthood.
FIXED AND RANDOM EFFECTS IN HLM. Fixed effects produce constant impact on DV. Random effects produce variable impact on DV. F IXED VS RANDOM EFFECTS.
Repair Dr Heyam Awad FRCPath. Tissue repair Restoration of tissue architecture and function after injury. Two types : 1) regeneration. 2) scar formation.
Shortest Path Analysis and 2nd-Order Analysis Ming-Chih Kao U of M Medical School
Eigengenes as biological signatures Dr. Habil Zare, PhD PI of Oncinfo Lab Assistant Professor, Department of Computer Science Texas State University 5.
Statistical Analysis for Expression Experiments Heather Adams BeeSpace Doctoral Forum Thursday May 21, 2009.
Protein. Protein and Roles 1: biological process unknown 1.1 Structural categories 1.2 organism categories 1.3 cellular component o unlocalized.
Research Methodology Proposal Prepared by: Norhasmizawati Ibrahim (813750)
Eigengenes as biological signatures Dr. Habil Zare, PhD PI of Oncinfo Lab Assistant Professor, Department of Computer Science Texas State University 3.
Comparison of Mouse Data to Human Datasets 3/1/16.
A New Statistical Method for Analyzing Longitudinal Multifactor Expression Data and It ’ s Application to Time Course Burn Data Baiyu Zhou Department of.
Integrin signalling Vytášek 2010.
Metabolism.
Volume 2, Issue 2, Pages (February 2016)
Integrin signalling Vytášek 2009.
Apoptosis Begins when a cell receives a “death signal”
Cell to Cell Communication
Fixed, Random and Mixed effects
Volume 3, Issue 1, Pages (July 2016)
Fig. 2. Pathways differentially regulated in patients with early Lyme disease and STARI. Pathways differentially regulated in patients with early Lyme.
Michal Levin, Tamar Hashimshony, Florian Wagner, Itai Yanai 
Chapter 10 Introduction to the Analysis of Variance
Kanyan Xu, Xiangzhong Zheng, Amita Sehgal  Cell Metabolism 
Inferring Cellular Processes from Coexpressing Genes
Chapter 22 – Comparing Two Proportions
Presentation transcript:

A Method for Analyzing Time Course Multi-factor Expression Data with Applications to A Burn Study Baiyu Zhou Department of Statistics Stanford University 12/08/2008

Outline Motivations Brief review: methodology Tissue data analysis Age impact on survival in adult burn patients

Data: gene expression at different times after burn injury Tissue data (blood, skin, muscle and fat): Questions: 1. what tissue is most affected by burn injury? 2. what tissue contributes most to pediatric and adult differences in burn patients? 3. For a given gene, how is its expression affected in different tissues? Age impact on survival in burn patients: Reported dramatically increasing death rate in burn patients over 48 years old ( Muller MJ, Pegg SP, Rule MR. Determinants of death following burn injury. Br J Surg Apr;88(4): ) The explanation on gene express level? Data and Questions These questions can be addressed by our method !

Brief Review: Methodology (1) A novel approach to analyze longitudinal multi-factorial data Example: time course measurements of gene expressions from each patients. longitudinal: time course multi-factorial: factor 1: burn/control ; factor 2: gender (male/ female) Classify genes into different gene sets. Each gene set represents a different ANOVA structure. Example: C1: interaction. Gender related burn responsive C2: additive. Burn and gender effects are independent C3: only have burn effect, no gender differences C4: only have gender differences, no burn effect C5: constant genes

Brief Review: Methodology (2) Gene classification is based on information pooling from time course measurements. Each gene is associated with an ANOVA direction (in time space), which captures gene specific response features. Response timing (when to respond): The magnitude of each component of the ANOVA direction reflects the intensity of ANOVA signal at the corresponding time points. Response pattern (how to respond): Negative signs in the ANOVA directions indicate the ANOVA signal is embedded into the expression change between time points. Examples (Burn+Age, time course data)

Application beyond time course data : Tissue Data Analysis Tissue data : Gene expressions in blood, skin, muscle and fat were measured for burn patients and controls. Two factors : burn/control and age (children ( =22 yr)). Questions: Which tissue is mostly affected by burn injury? Which tissue is perturbed most differentially in children and adults? For a gene of interest, how is its expression affected in different tissues? Analysis longitudinal multi-factorial data Treat tissue measurements as a vector Estimate ANOVA directions in ‘ tissue space ’ BurnControl Children79 Adults114 Data :

Tissue Data Analysis (1) The total of probe sets are classified into five gene sets (FDR=0.05). C1C2C3C Examples: A gene ’ s ANOVA direction reflects the ANOVA signal intensity in different tissues of that gene

Tissue Data Analysis (2) Question: Which tissue contributes most to age related differences in burn patients? Weight matrix First eigen-gene (first right singular vector of W) captures weight distribution among tissues for a gene set.

Tissue Data Analysis (3) Age differences are strong in muscle, fat and skin after burn injury (C1 and C2 genes) Gene expressions in blood are perturbed most by burn injury (C3) Blood Muscle Fat Skin Weight Matrix (C4) 01 Row Z-Score Color Key

Tissue Data Analysis (4) A gene is assigned to a tissue if the tissue receives the largest weight in the ANOVA direction. The gene sets are divided into four tissue subsets. TissueFunctional groups (GO: BP_5)pathways C3 (5205 probe sets) Blood (2242)Apoptosis (8.7E-8) Cell death (3.7E-7) Protein kinase cascade (4.5E-6) Natural killer cell mediated cytotoxicity (8.0E-4) GnRH signaling pathway (1.3E-3) T cell receptor signaling pathway (1.3E-3) Skin (1513)Vasculature development (4.2E-7) Blood vessel development (9.1E-7) Blood vessel morphogenesis (1.8E-5) Organ morphogenesis (8.4E-5) Focal adhesion (2.1E-5) Regulation of actin cytoskeleton (4.9E-4) Muscle (799)Cellular protein catabolic process (1.1E-3) Modification-dependant macromolecule catabolic process (2.4E-3) Protein catabolic process (2.9E-3) Tight junction (1.9E-2) Valine, leucine and isolecine degradation (1.9E-2) Fat (651)Tissue development (9.3E-3) Icosanoid metabolic process (2.7E-2) Fatty acid metabolic process (4.4E-2) Valine, leucine and isolecine degradation (3.6E-3) Fatty acid metabolism (4.0E-3) Sulfer metabolism (8.9E-3) Example: functional analysis on tissue subsets of C3.

Extension to Cross-sectional Time Course Data Longitudinal: time course measurements are from the same experimental units Cross-sectional : time course measurements are not from the same experimental units. Different numbers of measurements are allowed at each time points. Method for cross-sectional data: Assume gene specific variance. No correlations over time points. Estimate mean vectors for conditions Estimate ANOVA directions based on s Estimate gene specific variance by pooling information from all arrays

An Example : Impact of Age on Survival After Burn Injury One factor : Died/Survived Three time points (age groups) Two data sets: (1) early stage data (2) middle stage data Question: The impact of age on survival after burn injury. The data are cross-sectional Age group 1 ( year-old) Age group 2 ( year-old) Age group 3 (>=55 year-old) Data set 1 (early stage) Died8610 Survived Data set 2 (middle stage) Died757 Survived38267

Impact of Age on Survival After Burn Injury (1) 1236 (early stage ) and 1171 (middle stage) probe sets are differentially expressed between survival & non survival populations (FDR=0.05). Early Stage Data

Impact of Age on Survival After Burn Injury (2) Middle Stage Data

Impact of Age on Survival After Burn Injury (3) Weight matrix: Weight matrix reflects the impact of age groups for each gene. The first right singular vector of W ( ‘ first eigen-gene ’ ) reflects the impact of age on the gene set. Early data: age group 2 (40-54 yo) Middle data: age group 2 (40-54 yo) and 3 (>=55 yo)

Impact of Age on Survival After Burn Injury (4) The result coincide with the reported increasing death rate in burn patients over 48 year-old (7.3 time more likely to die after burn injury) Some significant pathways: Early Stage Data Middle Stage Data

Summary A novel approach for analyzing time course multi-factor expression data (1) Classify genes into different gene sets based on factor effects, suited for explorative study (2) The estimated ANOVA directions capture gene specific response features: response timing and response pattern Applications (1) Burn + Age (pediatric/adult), time course data (2) Burn + Gender, time course data (3) Burn + Tissue (early stage) (4) Age impact on adult survival (early stage & middle stage) (5) Survival + Gender, time course data Results are available:

Acknowledgements Professor Wing Wong Weihong Xu, Wenzhong Xiao from Davis lab

Thanks!