Copy Number Variation Methods and Data

Similar documents
Joint Modelling Approaches in diabetes research. Francisco Gude Clinical Epidemiology Unit, Hospital Clínico Universitario de Santiago

International Journal of Emerging Technologies in Computational and Applied Sciences (IJETCAS)

Parameter Estimates of a Random Regression Test Day Model for First Three Lactation Somatic Cell Scores

Study and Comparison of Various Techniques of Image Edge Detection

INTEGRATIVE NETWORK ANALYSIS TO IDENTIFY ABERRANT PATHWAY NETWORKS IN OVARIAN CANCER

THIS IS AN OFFICIAL NH DHHS HEALTH ALERT

The Limits of Individual Identification from Sample Allele Frequencies: Theory and Statistical Analysis

This article appeared in a journal published by Elsevier. The attached copy is furnished to the author for internal non-commercial research and

Modeling Multi Layer Feed-forward Neural. Network Model on the Influence of Hypertension. and Diabetes Mellitus on Family History of

Incorrect Beliefs. Overconfidence. Types of Overconfidence. Outline. Overprecision 4/22/2015. Econ 1820: Behavioral Economics Mark Dean Spring 2015

The Effect of Fish Farmers Association on Technical Efficiency: An Application of Propensity Score Matching Analysis

THE NATURAL HISTORY AND THE EFFECT OF PIVMECILLINAM IN LOWER URINARY TRACT INFECTION.

Insights in Genetics and Genomics

An Introduction to Modern Measurement Theory

Optimal Planning of Charging Station for Phased Electric Vehicle *

A Geometric Approach To Fully Automatic Chromosome Segmentation

Introduction ORIGINAL RESEARCH

Fast Algorithm for Vectorcardiogram and Interbeat Intervals Analysis: Application for Premature Ventricular Contractions Classification

Resampling Methods for the Area Under the ROC Curve

NUMERICAL COMPARISONS OF BIOASSAY METHODS IN ESTIMATING LC50 TIANHONG ZHOU

Physical Model for the Evolution of the Genetic Code

ARTICLE IN PRESS Neuropsychologia xxx (2010) xxx xxx

Statistically Weighted Voting Analysis of Microarrays for Molecular Pattern Selection and Discovery Cancer Genotypes

Sparse Representation of HCP Grayordinate Data Reveals. Novel Functional Architecture of Cerebral Cortex

A MIXTURE OF EXPERTS FOR CATARACT DIAGNOSIS IN HOSPITAL SCREENING DATA

Estimating the distribution of the window period for recent HIV infections: A comparison of statistical methods

HIV/AIDS-related Expectations and Risky Sexual Behavior in Malawi

HIV/AIDS-related Expectations and Risky Sexual Behavior in Malawi

Unobserved Heterogeneity and the Statistical Analysis of Highway Accident Data

SMALL AREA CLUSTERING OF CASES OF PNEUMOCOCCAL BACTEREMIA.

Using a Wavelet Representation for Classification of Movement in Bed

The Case for Selection at CCR5-D32

Subject-Adaptive Real-Time Sleep Stage Classification Based on Conditional Random Field

IMPROVING THE EFFICIENCY OF BIOMARKER IDENTIFICATION USING BIOLOGICAL KNOWLEDGE

Using Past Queries for Resource Selection in Distributed Information Retrieval

Project title: Mathematical Models of Fish Populations in Marine Reserves

A GEOGRAPHICAL AND STATISTICAL ANALYSIS OF LEUKEMIA DEATHS RELATING TO NUCLEAR POWER PLANTS. Whitney Thompson, Sarah McGinnis, Darius McDaniel,

Saeed Ghanbari, Seyyed Mohammad Taghi Ayatollahi*, Najaf Zare

Research Article Statistical Analysis of Haralick Texture Features to Discriminate Lung Abnormalities

BIOSTATISTICS. Lecture 1 Data Presentation and Descriptive Statistics. dr. Petr Nazarov

Richard Williams Notre Dame Sociology Meetings of the European Survey Research Association Ljubljana,

Estimation for Pavement Performance Curve based on Kyoto Model : A Case Study for Highway in the State of Sao Paulo

Balanced Query Methods for Improving OCR-Based Retrieval

Appendix F: The Grant Impact for SBIR Mills

INITIAL ANALYSIS OF AWS-OBSERVED TEMPERATURE

Key words: carcass, fertility, genotype-by-environment, liver fluke, milk, reaction norm

Incorporating prior biological knowledge for network-based differential gene expression analysis using differentially weighted graphical LASSO

Survival Rate of Patients of Ovarian Cancer: Rough Set Approach

Investigation of zinc oxide thin film by spectroscopic ellipsometry

310 Int'l Conf. Par. and Dist. Proc. Tech. and Appl. PDPTA'16

CONSTRUCTION OF STOCHASTIC MODEL FOR TIME TO DENGUE VIRUS TRANSMISSION WITH EXPONENTIAL DISTRIBUTION

A comparison of statistical methods in interrupted time series analysis to estimate an intervention effect

TOPICS IN HEALTH ECONOMETRICS

Price linkages in value chains: methodology

I I I I I I I I I I I I 60

Integration of sensory information within touch and across modalities

Evaluation of two release operations at Bonneville Dam on the smolt-to-adult survival of Spring Creek National Fish Hatchery fall Chinook salmon

Estimation of Relative Survival Based on Cancer Registry Data

Association between cholesterol and cardiac parameters.

Sequential meta-analysis to determine whether or not to start another trial: the high frequency versus conventional mechanical ventilation example

4.2 Scheduling to Minimize Maximum Lateness

BIOSTATISTICS. Lecture 1 Data Presentation and Descriptive Statistics. dr. Petr Nazarov

Hierarchical Prediction Errors in Midbrain and Basal Forebrain during Sensory Learning

A Meta-Analysis of the Effect of Education on Social Capital

Statistical Analysis on Infectious Diseases in Dubai, UAE

Analysis of the QRS Complex for Apnea-Bradycardia Characterization in Preterm Infants

Arithmetic Average: Sum of all precipitation values divided by the number of stations 1 n

Non-parametric Survival Analysis for Breast Cancer Using nonmedical

WHO S ASSESSMENT OF HEALTH CARE INDUSTRY PERFORMANCE: RATING THE RANKINGS

THE NORMAL DISTRIBUTION AND Z-SCORES COMMON CORE ALGEBRA II

Using the Perpendicular Distance to the Nearest Fracture as a Proxy for Conventional Fracture Spacing Measures

AUTOMATED DETECTION OF HARD EXUDATES IN FUNDUS IMAGES USING IMPROVED OTSU THRESHOLDING AND SVM

Reconstruction of gene regulatory network of colon cancer using information theoretic approach

Lymphoma Cancer Classification Using Genetic Programming with SNR Features

Impact of Imputation of Missing Data on Estimation of Survival Rates: An Example in Breast Cancer

AN ENHANCED GAGS BASED MTSVSL LEARNING TECHNIQUE FOR CANCER MOLECULAR PATTERN PREDICTION OF CANCER CLASSIFICATION

Economic crisis and follow-up of the conditions that define metabolic syndrome in a cohort of Catalonia,

Proceedings of the 6th WSEAS Int. Conf. on EVOLUTIONARY COMPUTING, Lisbon, Portugal, June 16-18, 2005 (pp )

Biomarker Selection from Gene Expression Data for Tumour Categorization Using Bat Algorithm

ASSESSMENT OF PARAMETRIC AND NON-PARAMETRIC METHODS FOR SELECTING STABLE AND ADAPTED SPRING BREAD WHEAT GENOTYPES IN MULTI - ENVIRONMENTS ABSTRACT

A New Machine Learning Algorithm for Breast and Pectoral Muscle Segmentation

PSI Tuberculosis Health Impact Estimation Model. Warren Stevens and David Jeffries Research & Metrics, Population Services International

Are Drinkers Prone to Engage in Risky Sexual Behaviors?

Prediction of Human Disease-Related Gene Clusters by Clustering Analysis

Kim M Iburg Joshua A Salomon Ajay Tandon Christopher JL Murray. Global Programme on Evidence for Health Policy Discussion Paper No.

NHS Outcomes Framework

TOBIT REGRESSION AND CENSORED CYTOKINE DATA. Terrence L. O Day. BS, St Vincent College, MBA, University of Phoenix, 2003

Gene Selection Based on Mutual Information for the Classification of Multi-class Cancer

USING DIFFERENTIAL GEOMETRIC LARS ALGORITHM TO STUDY THE EXPRESSION PROFILE OF A SAMPLE OF PATIENTS WITH LATEX-FRUIT SYNDROME

Does reporting heterogeneity bias the measurement of health disparities?

Semantics and image content integration for pulmonary nodule interpretation in thoracic computed tomography

National Polyp Study data: evidence for regression of adenomas

UNIVERISTY OF KWAZULU-NATAL, PIETERMARITZBURG SCHOOL OF MATHEMATICS, STATISTICS AND COMPUTER SCIENCE

A Linear Regression Model to Detect User Emotion for Touch Input Interactive Systems

Figure S1. 1g tumors (weeks) ikras. Lean Obese. Lean Obese 25 KPC

Statistical models for predicting number of involved nodes in breast cancer patients

Nonstandard Machine Learning Algorithms for Microarray Data Mining. Byoung-Tak Zhang

Alma Mater Studiorum Università di Bologna DOTTORATO DI RICERCA IN METODOLOGIA STATISTICA PER LA RICERCA SCIENTIFICA

Disease Mapping for Stomach Cancer in Libya Based on Besag York Mollié (BYM) Model

Desperation or Desire? The Role of Risk Aversion in Marriage. Christy Spivey, Ph.D. * forthcoming, Economic Inquiry. Abstract

Transcription:

Copy Number Varaton Methods and Data

Copy number varaton (CNV) Reference Sequence ACCTGCAATGAT TAAGCCCGGG TTGCAACGTTAGGCA Populaton ACCTGCAATGAT TAAGCCCGGG TTGCAACGTTAGGCA ACCTGCAATGAT TTGCAACGTTAGGCA ACCTGCAATGAT ACCTGCAATGAT TAAGCCCGGG TTGCAACGTTAGGCA Copy number varatons (CNVs) are regons >1kb n a genome that occur n dfferent copy number n a populaton.

CNVs n Cancer Cells Development of sold tumors s assocated wth acquston of complex genetc alteratons. These alteratons can be of any length, ncludng full chromosomes. Changes can be caused by underlyng falures n mantenance of genetc stablty, Canges may be promoted by postve selecton as they provde growth advantages. Regons commonly amplfed may contan genes that mprove the vablty and growth of the tumor cells.

CNVs n populatons CNVs rangng from 1kb - several Mb segregate n humans and model organsms. Medan length of known CNVs s ~100kb, but that number s shaky. CNVs often have lmted phenotypc mpact CNV-alleles are nherted n the germlne. >10% of the human genome has varable copy number. The genome of two ndvduals has an average dfference n length of ~10 Mb. CNVs cover genes (Redon et al(06): 2900 genes are covered by CNVs)

More CNV facts Most CNVs are sngletons In most genomc regons, CNV-mutatons are rare. CNVs are usually n hgh LD wth SNPs. 25-fold enrchment near segmental duplcatons. CNVs are generated by several types of events, e.g. non-allelc homologous recombnaton and retrotransposton. More detected CNVs are duplcatons, but t s not clear f ths s detecton bas.

CNVs are dstrbuted throughout the Genome

CNVs and dseases Ads: CNV coverng CCL3L1, large copy number s protectve of HIV/AIDS nfecton. Assocaton wth Crohn s dsease and BMI. Autsm: De novo deletons more common n patents wth Autsm. New Syndrome: Deleton on chromosome 17 defnes novel genetc dsorder.

How to fnd CNVs? Non-Mendelan Inhertance

How to fnd CNVs? Compettve Genomc Hybrdsaton Feuk et al. Nature Revews Genetcs 7, 85 97 (February 2006)

How to fnd CNVs? SNP assays rs2613837 1.40 1.20 Norm Intensty (B) 1 0.80 0.60 0.40 0.20 0 0 0.20 0.40 0.60 0.80 1 1.20 1.40 1.60 1.80 Norm Intensty (A)

How to fnd CNVs? Pared end resequencng

Method Matters

Challenges of the data analyss The sgnal s nosy, hybrdzaton ntenstes depend on many envronmental factors. In populaton samples, probes coverng CNVs are rare, about 1% of all probes cover the rare allele of a CNV. The state space s not well-defned. Especally n cancer, we may observe a large amplfcaton n copy number.

2 Algorthms CBS -Crcular Bnary Segmentaton HMM - Hdden Markov Model Basc dea: Consecutve probes are lkely to have the same copy number, thus the data s locally correlated.

Crcularly Bnary Segmentaton Ohlsen et al. Bostatstcs (2004), 5, 4, pp. 557 572 Method to analyze CGH data. Desgned for Cancer cell analyss-needs to model complex changes n copy number. Implemented n the program DNAcopy, whch s wdely used. Has a nce vsual output.

Change pont method Defnton: Let X 1, X 2,...X n be a sequence of random varables. An ndex ν s called a change-pont f X 1,..., X ν have a common dstrbuton functon F 0 and X ν+1,... have a dfferent common dstrbuton functon F 1 untl the next change-pont (f one exsts). Here, a change ponts represent the change n copy number along the genome.

Testng for the exstence of change ponts Let S = X 1 + + X, 1 n, be the partal sums. When the data are normally dstrbuted wth a known varance 2, the statstc for testng the null hypothess of no change aganst the alternatve of exactly one change at an unknown locaton s gven by Z B = max 1<n Z, where Z s a t-statstc for unequal sample szes and n- are estmates of the varance from data ponts 1,, and +1,,n. Crtcal values for ths test can be derved by Monte Carlo smulaton. The locaton of the change-pont s estmated to be such that Z B = Z.. 1 1 2 ˆ 1) ( ˆ 1) ( 1 2 2 n S S S n n n Z n n

Fndng more change ponts Assume v s a changepont. To dentfy addtonal change ponts, run the test on X 1,,X v and on X v+1,,x n. Repeated recursvely untl no further change pont can be detected. Ths algorthm does result n a multple testng problem as the probablty of fndng spurous change-ponts s a functon of the number of true change-ponts. Snce the true number of change-ponts s unknown no correcton s performed.

However Snce the bnary segmentaton procedure s based on a test to detect a sngle change, a potental problem wth t s that t cannot detect a small changed segment bured n the mddle of a large segment. Ths problem wth the bnary segmentaton procedure s due to the fact that t looks for only one change-pont at a tme. Segment 1 Segment 2

Crculary Bnary Segmentaton 3 segments 2segments

Crcular bnary segmentaton The test statstc s then Z C = max 1<jn Z j wth Note that Z C allows for both a sngle change( j = n) and two changes ( j < n). Once the null hypothess s rejected the change-pont(s) s (are) estmated to be (and j ) such that Z C = Z j Multple pars of change ponts can be detected wth the same recursve algorthm descrbed before. j n S S S j S S j n j n j n j Z j n j j n j j 1 2 2 1 1 2 ˆ 1) ( ˆ 1) (

Applcaton to data

HMM Frdlyand et al. Journal of Multvarate Analyss 90 (2004) 132 153 Goals: Identfy the number of states present n the data. Estmate the state at each probe s l.

Markov Chan Markov Chans are statstcal processes where each next state only depends on the present state. In a hdden Markov model, the actual states are unobserved, we only get ndrect sgnals.

Markov Chan Markov Chans are statstcal processes where each next state only depends on the present state. In a hdden Markov model, the actual states are unobserved, we only get ndrect sgnals.

CNVs as HMM 0 0 1 1 1 0 2 2 0 0 Probes coverng a CNV are locally correlated, a probe n a CNV s more lkely to be next to an other probe n a CNV.

HMM A HMM wth the fxed number of hdden states K can be characterzed n terms of three parameters: () the ntal state probabltes, ; () the transton probablty matrx, A () the collecton of Gaussan emsson probablty functons defned wthn each state, B. For HMMs, there are effcent EM-based algorthms to create maxmum lkelhood estmates of A, B and. Based on A, B and, ML- estmates for copy number status at each poston are generated.

HMM for CNVs Emsson dstrbutons are generally generated from outsde nformaton. Genotype nformaton can be analyzed as orthologous sgnal. Transton matrx provdes pror frequency of CNVs and pror length dstrbuton.

Comparng the methods Smulaton study by takng hybrdzaton ntenstes from a prmary breast tumor sample, assgnng the true CNV status dependent on the mean sgnal. CNV lengths followed the dstrbuton of stretches of hgh and low sgnal n the data. Gaussan nose was added. For each smulated dataset, breakponts were estmated usng three methods, DNAcopy, HMM and Glad. Successve CNVs were merged.

Comparson of Methods HMM had the greatest power to detect the shortest segments DNAcopy surpassng HMM for longer segments. DNAcopy had by far the lowest FDR for all segment lengths.

Comparng tagsnps and callng methods P(C) (freq. of mnor CNV allele) P(O A) (false postve rate) P(O C) (senstvty) 0.02 0.05 0.1 0.2 IF r 2 IF r 2 IF r 2 IF r 2 0.9 1.74 0.57 1.37 0.73 1.25 0.80 1.20 0.83 0.01 0.8 2.05 0.49 1.59 0.63 1.44 0.69 1.40 0.71 0.7 2.49 0.40 1.88 0.53 1.70 0.59 1.66 0.60 0.9 4.41 0.23 2.45 0.41 1.80 0.56 1.48 0.67 0.05 0.8 5.51 0.18 2.99 0.33 2.16 0.46 1.78 0.56 0.7 7.13 0.14 3.77 0.27 2.68 0.37 2.18 0.46 IF - Inflaton factor to overcome callng error.