The Complexities of Data Analysis in Human Genetics Marylyn DeRiggi Ritchie, Ph.D. Center for Human Genetics Research Vanderbilt University Nashville, TN

Biology is complex BioCarta

Single nucleotide polymorphisms (SNPs)

Mendelian Traits Aa BBbb AA aa BBBbbb Aa AAAa BBbbBb Locus 1 Locus 2 AABBAABbAAbb AaBBAaBbAabb aaBBaaBbaabb affected

Complex Traits Aa BBBb AA aa BBBbbb aa AAAa BBbbBb Locus 1 Locus 2 AABBAABbAAbb AaBBAaBbAabb aaBBaaBbaabb affected

Complex Traits Complex trait implies the involvement of multiple genes and/or environmental factors Mendelian trait implies a single mutation Mendelian traits are generally rare Complex traits are common and of substantial public health impact

Genetic Analysis Two main areas of genetic analysis 1.Linkage analysis 2.Association analysis Methods have been developed for each approach for a variety of different study designs

Association Analysis In disease studies, when the disease gene is unknown, we look for association between genetic markers and the disease If a marker occurs more frequently or less frequently in affected individuals than in unaffected individuals, then it is associated with the disease.

Association Analysis Case-control studies –Test for association between marker alleles and the disease phenotype in a group of affected and unaffected individuals randomly from the population Family-based studies –Test for association between marker alleles and the disease phenotype in a group of affected individuals and unaffected family members

Case-control data structure StatusSNP1SNP2SNP3SNP4SNP5SNP6SNP7SNP8SNP9SNP

Association Analysis Single marker tests Haplotype association Epistasis

Single marker tests SNP 1  Disease ??? SNP 2 SNP 3

Haplotype

Haplotype Analysis May be able to increase power by testing for association with marker haplotype Haplotype is a block of DNA that stays intact through generations Do not directly observe marker haplotypes Use likelihood methods to infer

Haplotype Analysis

Epistasis: Gene-Gene Interactions W. Bateson, Mendel’s Principles of Heredity (1909) A.R. Templeton, In: Wade et al. (eds), Epistasis and the Evolutionary Process (2000) Epistasis first used by William Bateson (1909) Literal translation is “standing upon” (I.e. one gene masks the effects of another gene). Genotype at Locus A Genotype at Locus B BBBbbb AA WhiteGrey Aa BlackGrey Aa BlackGrey Cordell, Human Molecular Genetics 11: (2002)

Gene-gene Interactions Searching for gene-gene interactions brings about a whole new suite of problems and challenges Types of interactions –Additive –Multiplicative –Epistatic Curse of dimensionality – big problem

Curse of Dimensionality AAAaaa SNP 1 N = Cases, 50 Controls

SNP 2 AAAaaa BB Bb bb N = Cases, 50 Controls SNP 1 Curse of Dimensionality

N = Cases, 50 Controls AA Aaaa BB Bb bb CCCccc DD Dd dd AA Aaaa AA Aaaa BB Bb bb BB Bb bb SNP 1 SNP 2 SNP 4 SNP 3 Curse of Dimensionality

Three Other Issues to Consider 1.Variable selection 2.Model selection 3.Interpretation

1. Variable Selection How can you determine which variables to select? Not computationally feasible to evaluate all possible combinations Need to select correct variables to detect interactions

How many combinations are there? ~500,000 SNPs span 80% of common variation in genome (HapMap) SNPs in each subset 5 x x x x x Number of Possible Combinations

How many combinations are there? ~500,000 SNPs span 80% of common variation in genome (HapMap) SNPs in each subset 5 x x x x x Number of Possible Combinations 2 x combinations * 1 combination per second * seconds per day x days to complete ( x years)

2. Model Selection For each variable subset, evaluate a statistical model Goal is to identify the best subset of variables that compose the best model

Finding the best model Choose variable subset Choose statistical model Evaluate model fitness Best model

Simple Fitness Landscape Model Fitness

Complex Fitness Landscape Fitness Model

3. Interpretation Selection of best statistical model in a vast search space of possible models Statistical or computational model may not translate into biology May not be able to identify prevention or treatment strategies directly Wet lab experiments will be necessary, but may not be sufficient

3. Interpretation Strategies to assess biological interpretation of gene-gene interaction models 1.Consider current knowledge about the biochemistry of the system and the biological plausibility of the models 2.Perform experiments in the wet lab to measure the effect of small perturbations to the system 3.Computer simulation algorithms to model biochemical systems

Additional Challenges (true of all association studies) Sample size and power/type I error Population specific effects –Age, gender Poorly matched cases and controls –Ethnic background –Controls must be “at risk” Bias Heterogeneity

Phenotypic (Clinical, Trait) –Affected individuals vary in clinical expression Genetic –Different inheritance patterns for same disease Locus –Different genes lead to the same disease Allelic –Different alleles at the same gene lead to same/different disease Thornton-Wells TA, Moore JH, Haines JL. Trends in Genetics, 2004;20(12):

New Statistical Approaches Data Reduction –Combinatorial Partitioning Method (CPM) –Multifactor Dimensionality Reduction (MDR) –Detection of informative combined effects (DICE) –Logic Regression –Set Association Analysis Pattern Recognition –Symbolic Discriminant Analysis (SDA) –Cellular Automata (CA) –Neural Networks (NN)

Areas of Future Work (possible collaborations) More analytical methods for gene-gene and gene-environment interactions –Especially including categorical and continuous variables simultaneously Inclusion of pathway information into analyses Ways of dealing with heterogeneity of all kinds