GENOME-WIDE GENETIC INTERACTION ANALYSIS OF GLAUCOMAUSING EXPERT KNOWLEDGE DERIVED FROM HUMAN PHENOTYPE

NETWORKS

TING HU†, CHRISTIAN DARABOS†, MARIA E. CRICCO, EMILY KONG, JASON H. MOORE

Institute for the Quantitative Biomedical Sciences, Geisel School of Medicine, Dartmouth CollegeHanover, NH 03755, U.S.A.

E-mail: jason.h.moore@dartmouth.edu

The large volume of GWAS data poses great computational challenges for analyzing geneticinteractions associated with common human diseases. We propose a computational framework forcharacterizing epistatic interactions among large sets of genetic attributes in GWAS data. We buildthe human phenotype network (HPN) and focus around a disease of interest. In this study, we usethe GLAUGEN glaucoma GWAS dataset and apply the HPN as a biological knowledge-based filterto prioritize genetic variants. Then, we use the statistical epistasis network (SEN) to identify asignificant connected network of pairwise epistatic interactions among the prioritized SNPs. Theseclearly highlight the complex genetic basis of glaucoma. Furthermore, we identify key SNPs byquantifying structural network characteristics. Through functional annotation of these key SNPsusing Biofilter, a software accessing multiple publicly available human genetic data sources, wefind supporting biomedical evidences linking glaucoma to an array of genetic diseases, proving ourconcept. We conclude by suggesting hypotheses for a better understanding of the disease.

Keywords: GWAS; Epistasis; Gene-gene interaction; Human phenotype network; Statistical epistasisnetwork; Biofilter; Eye diseases; Glaucoma; SNPs; Pathways.

1. Introduction

With the rapid development of genotyping technologies and exponential increase in computa-tional power, we are able to leverage the wealth of genome-wide association studies (GWAS)data to test millions of genetic variations (single nucleotide polymorphisms, SNPs) for theirassociations with common human diseases.1,2 However, common disorders are usually theresult of the non-linear combined effect of many variations. The study of these complex,epistatic interactions among multiple genetic attributes is crucial in explaining portions of thegenotype-to-phenotype associations.3,4

Detecting and analyzing complex genetic interactions in GWAS data pose great statisticaland computational challenges. Epistatic effects potentially involve any number of SNPs, froma couple to hundreds or even thousands. In turn, the uncertainty of size of the interacting SNPset would require interaction studies to comprehensively explore all possible combinations ofSNPs. A typical GWAS dataset encompasses hundreds of thousands of variations, makingthe enumeration of all pairwise SNP interactions computationally infeasible, and the com-putational cost increases exponentially with the number of features at hand. This inevitablybecomes a computational bottleneck of processing and analyzing GWAS data.

Therefore, genetic interaction studies meant to process modern GWAS data require thedevelopment of advanced informatics methodologies. These novel algorithms must be able to

†Co-first authors – TH and CD have contributed equally to this work.

simultaneously analyze large sets of interactions and identify genetic attributes potentiallyinvolved in higher-order interactions. Network modeling has emerged as a suitable frameworkfor such purposes,5–7 thanks to its ability to represent a large number of entities, as vertices,and their relationships, as edges.

In this study, we propose an informatics framework of detecting and analyzing genetic in-teractions for GWAS data. We use the example of the GLAUGEN study, a GWAS on glaucomaconsisting of over a million attributes. We pre-screen the dataset using a knowledge-based fil-ter by building human phenotype network (HPN) to prioritize SNPs and reduce the searchspace according to their known relationship to the disease/phenotype in question. This re-duced SNP list is then used to quantitatively evaluate all pairwise epistatic interactions andto build statistical epistasis network (SEN) to characterize their global interaction structure.The key SNPs identified by the SEN are functionally annotated and evaluated for their inter-action with other disorders. This new framework has the potential to elucidate large parts ofthe complex genetic architecture of common human diseases, such as glaucoma, and identifykey SNPs for further biological validations.

2. Dataset and Methods

2.1. Glaucoma and the GLAUGEN Study

Glaucoma, a neurodegenerative disease, is the primary cause of irreversible blindness, affectingover 60 million people worldwide. The most common kind of glaucoma, in all populations, isprimary open-angle glaucoma (POAG). Currently the only modifiable risk factor of POAGis intraocular pressure (IOP), but even lowering that will only slow the process, not stop it.8

However, a substantial amount of POAG has been shown to have a genetic basis. Familialaggregation of POAG has been long recognized and studied to find multiple loci linked tothem, causing the discovery of glaucoma-causing genes myocilin (MYOC), optineurin (OPTN)and WDa-repeat domain 36.9 About 5% of POAG is presently attributed to a single-gene orMendelian forms of glaucoma. More cases of POAG are caused by the combined epistaticeffects of many genetics risk factors.10

The Glaucoma Gene Environment (GLAUGEN) seeks to illuminate the origin of the dis-ease, to discover genetic loci associated with POAG, and to identify gene-gene and gene-environment associations.8 GLAUGEN is a GWAS, case-control study, with about 2,000 un-related cases and over 2,300 controls. To be involved in the study, both the cases and thecontrols had to be at least 40 years of age and European derived or Hispanic Caucasian.Subjects were genotyped with 1,048,965 SNPs examined.8

2.2. Human Phenotype Network (HPN)

The human disease network,11 or the more general human phenotype networks (HPNs),12,13

are mathematical graph models where nodes represent human genetic disorders and edgeslink those nodes with shared biology.14 The underlying connections of the HPN contribute to

atryptophan-aspartic acid

Permanent tooth development

Response to platinum-based chemotherapy in non-small-cell lung cancer

Ovarian reserve

beta Carotene

Heart Function Tests

Odontogenesis

Behcet's disease

Membranous Glomerulonephritis

Pulse

Serum uric acid levels

Response to TNF antagonist treatmentWalking

PR interval

Heart Atria

Alanine Transaminase

Tooth EruptionAcenocoumarol

Emphysema

Gallbladder cancer

Left Atrial Function

Brain structure

Lupus Vulgaris

Interleukin-1beta

Cyclic Peptides

bipolar disorder and depression (combined) Schizophrenia

Testicular Neoplasms

Parathyroid Hormone

1-Alkyl-2-acetylglycerophosphocholine Esterase

Electrocardiography

Sj�gren's syndrome

Lung cancer

Frontal Lobe

External Ear

Open-Angle Glaucoma

Dupuytren's disease

Exfoliation Syndrome

Phytosterols

Speech Perception

alpha-Macroglobulins

vWF and FVIII levels

Subcutaneous Fat

Glucose Tolerance Test

Adverse response to carbamapezine

Urinary Bladder Neoplasms

Hair

Paget's disease

Vascular endothelial growth factor levelsSelf-reported allergy

Primary biliary cirrhosis

5-HTT brain serotonin transporter levels

Bladder cancer

Coronary spasm

oxaliplatin [Supplementary Concept] Telomere length

Respiratory Function TestsBrain imaging

Menarche and menopause (age at onset) Cystatin CSepsis

Precursor Cell Lymphoblastic Leukemia-Lymphoma

Menarche

Nevus count

Citalopram

Carotid Artery Diseases

Bone Density

Atrial Natriuretic Factor

Lymphocytic Leukemia

Eye color

Ankle Brachial Index

Thyroid cancer

Idiopathic pulmonary fibrosis

Vitamin B 12

Common traits (Other)

Sleep

Forced Expiratory Volume

Aging traits

Depression

Gallbladder Neoplasms

White matter integrity

Esophageal cancer

Endometriosis

Ileal carcinoids

Major CVD

T-Lymphocytes

Atherosclerosis

Digit length ratio

Interleukin-6

Tumor biomarkers

Coronary Vasospasm

Carbamazepine

Biliary Liver CirrhosisBrain Mapping

Thyroid Neoplasms

Cytomegalovirus Vaccines

Soluble levels of adhesion molecules

Alcohol Drinking Sex Hormone-Binding Globulin

Thyrotoxic hypokalemic periodic paralysis

Quantitative traits

HIV-1 viral setpoint

Neurobehavioral Manifestations

Luteinizing Hormone

Glycoproteins

Pancreatic cancer

Optic Nerve

Antidepressive Agents

Urinary albumin excretion

Hodgkin Disease

CD8-Positive T-Lymphocytes

Partial Thromboplastin Time

Astigmatism

Myeloid Leukemia

Ferritins

Colorectal cancer

Primary Ovarian Insufficiency

Palmitic acid (16:0) plasma levels

Memory

Social Desirability

Apolipoproteins C

Non-albumin protein levels

Hoarding

Amygdala

Occipital Lobe

carbohydrate-deficient transferrin [Supplementary Concept]

Meningococcal disease

Aromatase Inhibitors

Nasopharyngeal Neoplasms

Cognitive test performance

Anemia

Cornea

Lymphoid Leukemia

Eosinophils

Visual Cortex

Benzodiazepines

Anthropometric traits

Metabolic Syndrome XEndometrial cancer

Perphenazine

Chemokines

Germinoma

Tumor Necrosis Factor-alpha

Sneezing

Cardiac Arrhythmias

Response to gemcitabine in pancreatic cancer

Movement Disorders

Select biomarker traits

Thoracic Aortic Aneurysm

Resting heart rate

Glycemic traits

Endometrial NeoplasmsCardiovascular Diseases

Tamoxifen

lipoprotein A-I [Supplementary Concept]

Intestinal Diseases

Maximal Midexpiratory Flow Rate

Extraversion (Psychology)

Pain

Ulcerative colitis

AIDS

Chronic Hepatitis B

Glucosylceramides

Dietary macronutrient intake

Carotid intima media thicknessDysplastic Nevus Syndrome

Migraine Disorders

Crohn's disease and celiac disease

Mean forced vital capacity from 2 exams

Tetralogy of Fallot

Celiac disease and Rheumatoid arthritis

Insulin-Like Growth Factor I

QT interval

Hyperkalemic Periodic Paralysis

Vital Capacity

Butyrylcholinesterase

Urinalysis

Cystatins

Beta-2 microglubulin plasma levels

gamma-Glutamyltransferase

Brain Natriuretic Peptide

Blood Glucose

Blue vs. brown eyes

Allopurinol

Complement C4b-Binding Protein

Reading and spelling

Proinsulin levels

Hyperactive-impulsive symptoms

Vascular Calcification

Obesity

Cortical thickness

Vitamin K

Oligospermia

Neurotic Disorders

Cardiovascular disease risk factors

Hormone measurements

Braces

Weight

Inflammatory biomarkers

Transforming Growth Factor beta1

Blood Platelets

Suntan

Gastric cancer

Cleft Lip

Interleukin-6 Receptors

T-tau

Personality dimensions

Uterine fibroids

Folic Acid

Intercellular Adhesion Molecule-1

L-Lactate Dehydrogenase

Skin pigmentation

Basal Cell Carcinoma

Word reading

Caffeine

Protein quantitative trait loci

Erectile Dysfunction

Follicle Stimulating Hormone

Primary sclerosing cholangitis

Isoxazoles

Breast Neoplasms

Morbidity-free survival

Insulin-Like Growth Factor Binding Protein 3

Interleukin-2 Receptor alpha Subunit

Migraine

Stearic acid (18:0) plasma levels

Folate pathway vitamin levels

Cardiac hypertrophy

Behavioural disinhibition (generation interaction)

Muscular Diseases

Monocyte count

Iris

Cervical cancer

Aorta

Chronic Kidney Failure

LDL Lipoproteins

Fuchs Endothelial Dystrophy

Hair Color

Polysaccharides

Hepatitis B Vaccines

Glaucoma

Corneal Topography

Eye

Alpha-Globulins

Hippocampus

Brain Waves

Optic Disk

Prostate cancer

Anti-cyclic Citrullinated Peptide Antibody

Intracranial volume

Reasoning

Two-hour glucose challenge

Inattentive symptoms

Circulating cell-free DNA

Floxacillin

Jaw Abnormalities

Cognitive decline

Papillomaviridae

Fibromyoma Myoglobin

Alopecia

Vitamin A

Central Nervous System

Vitamin B 6

Hippocampal volume

Apolipoproteins E

Acquired Immunodeficiency SyndromeAngiotensin-converting enzyme activity

Glomerulosclerosis

human [Supplementary Concept] F8 protein

Leptin

HIV-1 susceptibility

Blood trace element (Zn levels)

Chemokine CCL2

Serotonin Plasma Membrane Transport Proteins

MelanomaIGA Glomerulonephritis

InsulinEpirubicin-induced leukopenia

Lung Neoplasms

Allergic Rhinitis

Acetaminophen

Metabolic traits

Face

Aging

Neuroanatomy

SmokingDepressive Disorder

Transferrin

Blood pressure measurement (cold pressor test)

Inflammatory bowel disease

mean corpuscular hemoglobin concentration

Dengue shock syndrome

Autistic Disorder

Gallbladder Diseases

IgE levels

Colony-Forming Units Assay

alpha-Tocopherol

Red blood cell count

Chronic myeloid leukemia

Blood Flow Velocity

Esophagitis

Tissue Plasminogen Activator

Endothelial function traits

Response to acetaminophen (hepatotoxicity)

Smoking behavior

BrainNeutrophils

Common Carotid Artery

Temporal Lobe

Dengue Hemorrhagic Fever

Pure-Tone Audiometry

Abdominal Aorta

Taste

InflammationVenous thromboembolism

Antipsychotic Agents

Testicular cancer

Caffeine consumption

Cytomegalovirus antibody response

Intuition

Chronic hepatitis B infection

Migraine without aura

AsthmaKidney stones

Crohn Disease

Myeloproliferative Disorders

Follicule stimulating hormone

Liver Cirrhosis

Oocytes

Chemokine CCL4

Gout

Interleukin-12

Non-substance related behavioral disinhibition

Metabolism

Eye color traits

Transferrin Receptors

RisperidoneAddiction

Circadian Rhythm

Nasopharyngeal carcinoma

Adverse response to aromatase inhibitors

Subclinical atherosclerosis traits (other)

IgA levels

Sphingomyelins

Sudden Death

Osteitis Deformans

Creutzfeldt-Jakob Syndrome

End-stage coagulation

Rhabdomyolysis

Keratoconus

Cardiomegaly

Vascular Diseases

von Willebrand Factor

Neuropsychological Tests

Psychological Sexual Dysfunctions

Warfarin

Myopia

Soluble ICAM-1

Leukopenia

Stevens-Johnson Syndrome

Hearing Loss

Left Ventricular Function

Cardiac structure and function

Response to iloperidone treatment (PANSS-T score)

Autism

Prion diseases

Pancreatic Neoplasms

Alkaline Phosphatase

Anxiety

N-glycan levels

Multiple myeloma

Goiter

Event-related brain oscillations

Dialysis-related mortality

Allergic sensitization

Response to Vitamin E supplementation

Mental Disorders

Neuranatomic and neurocognitive phenotypes

Functional Laterality

Recombination rate (males)

Facial morphology

Foot

Hand Strength

Pulmonary function

Carotid Stenosis

Amyloid beta-Peptides

Response to platinum-based agents

Anthropometry

Religion and Psychology

Melanosis

Angiotensin-Converting Enzyme Inhibitors

Thyroid volume

Hodgkin's lymphoma

Retinal Vein

Interleukin-10

Echocardiographic traits

Androgen levelsTelomere

Response to cerivastatin

CerebrumSclerosis

Illicit drug useMyasthenia gravis

Heart rate variability traits

Gambling

Response to radiation

Immunoglobulin E

Thyrotropin

Mortality

Brugada syndromeCeliac disease

Blood Viscosity

Phospholipids

Mortality among heart failure patients

tau Proteins

Volumetric brain MRI

P-tau181p

Small Cell Lung Carcinoma

Adiponectin

Aortic root size

Systemic lupus erythematosus and Systemic sclerosis

Emphysema-related traits

gemcitabine [Supplementary Concept]

Albuminuria

Response to antineoplastic agents

Prostatic Neoplasms

Protein C

DehydroepiandrosteroneTanning

Kidney Diseases

Colonic Neoplasms

Working memory

Vascular Endothelial Growth Factor A

Brachial Artery

Oleic acid (18:1n-9) plasma levels

Essential Tremor

Iris characteristics

Caudate nucleus volume

Lipid Metabolism

Palmitoleic acid (16:1n-7) plasma levels

Hypothyroidism

Substance-Related Disorders

Gestational Diabetes

Paget's disease

Forced Vital Capacity

Angiography

Azoospermia

Hair morphology

Interleukin-18

Glioma

Tobacco Use Disorder

Head and Neck Neoplasms

Hematological and biochemical traits

Rheumatoid arthritis

Esophageal adenocarcinoma

Menopause

Focal Segmental Glomerulosclerosis

Hearing impairment

Drug-Induced Liver Injury

Iron deficiency

Esophageal Neoplasms

Interleukin-8

Psychomotor Performance

Blood Cells

Stomach Neoplasms

Personality

Sclerosing Cholangitis

Breast size

Metabolome

Anticonvulsants

Homocysteine

Attention Deficit Disorder with Hyperactivity

Serum selenium levels

Sunburns

Blood Sedimentation

Blood Vessels

Renal sinus fat

Clozapine

Macular Degeneration

Radiation response

Ovarian cancer

Autoimmune Diseases

Urinary metabolites

Short-Term Memory

Sensory disturbances after bilateral sagittal split ramus osteotomy

Colorectal Neoplasms

Thyroxine

Cleft Palate

Aspartate aminotransferase

Entorhinal Cortex

Acute lung injury

Vitamin K 1

Graves Disease Atrioventricular conduction

Vitamin E

Natriuretic peptide levels

Cortical structure

Erythrocyte Indices

Kawasaki disease

Nephropathy

DNA

radiation-related) Thyroid cancer (Papillary

Functional MRI

Kidney Calculi

Duodenal Ulcer

Coffee

Internal Carotid Artery

Pain Measurement

HIV-1

Handedness in dyslexia

Nevi and Melanomas

Migraine - clinic-based

Addictive Behavior

LongevityMonocyte early outgrowth colony forming units

Drug Toxicity

Vascular constriction

Chronic kidney disease

Insulin-Like Growth Factor Binding Protein 4

Parietal Lobe

Serum dimethylarginine levels (symmetric)

Comprehensive strength and appendicular lean mass

Aortic-valve calcification

Caudate Nucleus

Erythropoietin

Immunoglobulin A

Venous Thrombosis

Platelet function and related traits

Glucose Transporter Type 2

Biomedical quantitative traits

Exploratory Behavior

Soluble E-selectin levels

Marijuana Abuse

Mucocutaneous Lymph Node Syndrome

Osteoporosis

Attention Deficit and Disruptive Behavior Disorders

Insulin-Like Growth Factor Binding Protein 5

Antipsychotic-induced QTc interval prolongation

Asperger SyndromeLymphocytes

Sleep duration

Moyamoya disease

Survival

Infantile hypertrophic pyloric stenosis

E-Selectin

Carotid Arteries

Retinol levels

Fuchs's corneal dystrophy

gamma-Glutamylcyclotransferase

Ovarian NeoplasmsFructose-Bisphosphate Aldolase

Meningococcal Infections

Apolipoprotein A-I

Peroxidase

Intra-Abdominal Fat

Abdominal Fat

Coronary Disease

Vitamin D

Thyroid hormone levels

Blue vs. green eyes

Sphingolipids

Hypertrophic Pyloric Stenosis

Upper aerodigestive tract cancers

Response to tamoxifen in breast cancer

Life Expectancy

Ceramides

Serum alkaline phosphatase levels

β2-Glycoprotein I (β2-GPI) plasma levels

Progranulin levels

Insulin-like growth factors

Adverse response to lamotrigine and phenytoin

Dehydroepiandrosterone Sulfate

Dupuytren Contracture

Coronary restenosis

Abdominal Aortic Aneurysm

Male fertility

Leukocyte Count

fibrin fragment D [Supplementary Concept]

CD40 Ligand

Systemic Scleroderma

Apolipoprotein Levels

Chronic lymphocytic leukemia

Retinal vascular caliber

Behavior

Carcinoid Tumor

Heart Diseases

Interleukin 1 Receptor Antagonist ProteinBleomycin sensitivity

total Cholesterol

Iris color

Response to treatment for acute lymphoblastic leukemia

Corneal curvature

Iron-Regulatory Proteins

Coffee consumption

Pubertal anthropometrics

Blood Proteins

Barrett's esophagus

Treatment response for severe sepsis

Non-Small-Cell Lung Carcinoma

Genetic Recombination

Lymphoma

Migraine with aura

Digestive system disease (Barrett's esophagus and esophageal adenocarcinoma combined)

Haptoglobins

Blood Coagulation Factor Inhibitors

P-Selectin

Information processing speed

Circulating vasoactive peptide levels

Hepatocellular carcinoma

Type 2 diabetesChemerin levels

Hepatitis B vaccine response

0"

100"

200"

300"

1" 5" 9" 13" 17"

frequ

ency"

degree"

0.1$

1$

10$

100$

1000$

1$ 10$ 100$

Fig. 1. The SNP-based human phenotype network, filtered (edge weight cutoff = 0.002). Vertices are coloredby “modules”13 to increase readability. The vertex sizes are proportional to the number of associated SNPs.The degree distribution is given on both linear and logarithmic scales.

the understanding of the basis of disorders, which in turn leads to a better understanding ofhuman diseases.

In the present work, we rely on the SNP-based HPN, where connected diseases share com-mon SNPs. We use two sources for the phenotype-to-SNPs mapping: the National HumanGenome Research Institute GWAS catalog,15 a manually curated list of all the National Insti-tute of Heath (NIH) funded GWAS studies, and the NIH’s database of Genotypes and Phe-notypes (dbGaP, http://www.ncbi.nlm.nih.gov/gap). To build the HPN, all traits listed inthe GWAS catalog and dbGaP are annotated with their risk-associated SNPs. All the traitsbecome nodes in our network. Then, traits that have one or more SNPs in common are linkedin the HPN by an edge, the weight of which is proportional to the size of the SNPs’ overlap.The GWAS catalog and dbGaP report 1,252 traits combined, annotated with 37,681 SNPs in16,411 loci. The resulting bipartite network is projected in the space of phenotype vertices toobtain the HPN shown in Fig. 1.

The HPN contains 985 vertices and over 26,000 edges, and encompasses all phenotypeslisted in the GWAS catalog and dbGaP, provided that they are connected to at least one othertrait. The resulting network is extremely dense, with an average degree greater than 500.

2.3. Statistical Epistasis Network (SEN)

Statistical epistasis network (SEN) is designed to study a global aggregation of pairwiseepistatic interactions among a large number of factors.16,17 First, pairwise epistatic interaction

is quantified using the information-theoretic measure called information gain18 for all possiblepairs of genetic attributes. Specifically, the genotypes of two SNPs A and B, and the discretephenotypic class C are regarded as variables. Mutual information I(A;C) or I(B;C) measuresthe shared information between (A or B) and C calculated as I(A;C) = H(A)+H(C)−H(A,C)

or I(B;C) = H(B) +H(C)−H(B,C), where entropy H measures the uncertainty of a randomvariable or joint multiple random variables. Therefore, mutual information I(A;C) or I(B;C)

describes the reduction of the uncertainty of the phenotypic class C given the knowledge ofthe genotype of A or B, known as the main effect of A or B on C. Similarly, the joint mutualinformation I(A,B;C) measures the total effect of combining A and B on explaining the phe-notype C. Subtracting the individual mutual information I(A;C) and I(B;C) from I(A,B;C)

captures the gained information on C by considering A and B together rather than individ-ually. This information gain IG(A;B;C) is a practical measure for the epistatic interactioneffect of SNPs A and B on phenotype C, and has been used as an efficient non-parametric andmodel-free statistical quantification of pairwise epistasis in genetic association studies.19–23

Second, all pairwise SNP-SNP interactions are quantified and ranked. We build statisticalepistasis networks where vertices are SNPs and edges are weighted pairwise interactions. Weonly include pairs of SNPs if their interaction strengths are stronger than a preset threshold.We analyze the network topological properties at each cutoff value of the threshold, such as thesize of the network (the number of its vertices and the number of its edges), the connectivity ofthe network (the size of its largest connected component), and its vertex degree distribution.

Then, a threshold of the pairwise interaction strength is determined systematically byfinding the cutoff where the topological properties of the real-data network differentiate themost from the null distribution of permutation testing. Such a SEN provides a significantglobal structure of clustered strong pairwise epistatic interactions associated with a particularphenotype. It serves as a map for further network properties investigation and key SNPsprioritization as discussed in the following section.

2.4. Network Property Analysis of SEN

The assortativity of a network measures the propensities of vertices with similar characteristicsto connect to one another.24,25 In the context of SENs, we are interested in looking into themain effect assortativity, i.e., whether there exists a correlation of main effects between pairsof interacting SNPs. Such main effect assortativity is calculated as the Pearson correlationcoefficient r of the main effects at either ends of an edge in a SEN,

r =M−1

∑i jiki − [M−1

∑iji+ki

2 ]2

M−1∑

ij2i +k

2i

2 − [M−1∑

iji+ki

2 ]2, (1)

where M is the total number of edges, and ji and ki are the main effects, calculated asthe mutual information of a SNP and the phenotypic class of the vertices (SNPs) at theends of the i-th edge, with i = 1, 2, ...,M . The coefficient r lies between -1 and 1, with r =

1 indicating perfectly assortative, r = 0 for non-assortative, and r = −1 for complete dis-assortative networks.

At the vertex level, centrality measures the importance of individual vertices in a network.The most commonly used centrality measure is the node’s degree CD(v), which is the total

Phenotype #SNPs DegreeExfoliation Syndrome 1 1Coronary Artery Disease 639 2Cardiovascular Diseases 70 2Corneal curvature 13 4Optic Disk 17 4Open-Angle Glaucoma 6 5Glioma 12 2Eye 13 4Glaucoma 18 9Coronary restenosis 56 3

2SHQ�$QJOH�*ODXFRPD

([IROLDWLRQ�6\QGURPH

&DUGLRYDVFXODU�'LVHDVHV

*ODXFRPD (\H

2SWLF�'LVN

&RURQDU\�$UWHU\�'LVHDVH

*OLRPD

&RURQDU\�UHVWHQRVLV

&RUQHDO�FXUYDWXUH

(a) (b)

Fig. 2. Glaucoma network neighborhood. (a) Phenotypes, including the number of SNPs associated witheach phenotype and the “strength” of the interaction (degree) with the rest of the sub-network. (b) a depictedrepresentation of the Glaucoma-centered sub-network, in which the node sizes are proportional to the numberof associated SNPs.

number of edges connected to vertex v. The degree of a SNP in the statistical epistasis networkshows the number of other SNPs with which it is interacting. Betweenness centrality is a moresophisticated metric that quantifies the number of times a vertex is part of the shortest pathbetween any pair of vertices,26 represented as CB(v) =

∑s 6=v 6=t∈V

σst(v)σst

, where σst is the totalnumber of shortest paths from vertex s to vertex t and σst(v) is the number of those pathsthat pass through vertex v. Closeness centrality, defined as CC(v) = 1∑

s 6=v dvs, where dvs is the

distance between vertices v and s.27,28 This metric describes how easily a given vertex canreach all other vertices in a network. In the context of SENs, centrality measures are used toidentify key SNPs that play an essential role in the global interaction structure.

3. Results

3.1. Pre-screening of Glaucoma Data Using HPN

To reduce the computational cost (in time and resources) of analyzing a GWAS dataset, thedata must be filtered, selected and/or prioritized prior to processing. Algorithmic filteringmethods fall into two main families: knowledge-driven and data-driven. Knowledge-drivenfilters relies on the actual known biology behind the data, whereas data-driven filtering solelyrelies on the statistical pre-processing of the data, regardless of their type. ReliefF, SURF,and TuRF29,30 are examples of data-driven statistical methods of preprocessing GWAS data.For a complete review of the knowledge-based vs. data-driven filtering, refer to Sun et al.31

The method we propose to use here starts with the full HPN described in Section 2.2.We identify the glaucoma vertex and establish a list of the disorders and phenotypes in itsimmediate neighborhood. The 10 members of the “glaucoma neighborhood” are presented intable in Fig. 2.

Using the sub-HPN, we compile a list of 948 SNPs that have been associated with anyof the phenotypes in the sub-network. We subsequently annotate each of the risk-associatedSNPs with its linkage disequilibrium32 SNPs, indirectly associated with an increased risk. Thefull list of 4,388 features containing both direct and indirect risk-associated SNPs is usedto filter the full one million SNPs of the GLAUGEN GWAS data. The subset made of theoverlap between our priority-list and the GLAUGEN dataset, composed of 2,380 SNPs, is

used to build the SNP interaction network presented in the following section.

3.2. Statistical Epistasis Network of Glaucoma

Using the list of 2,380 SNPs, we evaluate all the pairwise epistatic interactions (>2.8 million)using the information gain measure. The strongest interacting pair is SNPs rs13123790 andrs6572380 with a strength of 1.163% association of the disease class. We set the pairwiseinteraction strength threshold decreasing from the strongest observed value with a step of0.0001. At each cutoff, a SEN is built by including pairs of SNPs as edges and two endvertices if their interaction strength is above the cutoff. Then for each network we evaluate itstopological properties including the number of non-isolated vertices, the number of edges, thesum of weighted edges, and the size of its largest connected component. A 100-fold permutationtest assesses the significance of these observed network properties. We permute the data 100times by randomly shuffling the disease class column, and for each permuted dataset all thepairwise interaction strengths are calculated and networks are built using the same cutoffs asthe read-data networks.

We observe a network at cutoff 0.693% that has a largest connected component includingsignificantly more vertices (p = 0.05) than those permuted-data networks at the same cutoff.This largest connected component has 713 vertices and 789 edges (Fig. 3).

The main effect assortativity of this network is -0.053 with a significance of p = 0.04

using a 100-fold edge swapping permutation test where we randomly pick 10× |E| (|E| is thetotal number of edges) pairs of edges and swap their end vertices. This significant negativeassortativity indicates that the main effects of interacting SNPs are negatively correlated, i.e.,high main-effect vertices tend to interact more with low main-effect vertices.

The degree, betweenness, and closeness centralities of all the vertices in the network areevaluated, and their distributions are shown in Fig. 4. Most of vertices have degrees equal toor less than 3. However there are a minority vertices with a degree higher than 5, meaning thatthey interact with larger number of other SNPs. In the distribution of betweenness centrality,the majority of vertices have low betweenness. However, some vertices have a much higherbetweenness than the rest. The closeness centrality follows a normal distribution indicatingthat most vertices have similar access to all other vertices. We sort the vertices by descendingcentrality. At the top, we find the key SNPs that are potentially involved the genetic processesresponsible for glaucoma. We further investigate the clinical and biomedical implications ofthese SNPs through Biofilter33 functional annotations. Biofilter provides a single interfaceto multiple publicly available online human genetic datasources (https://ritchielab.psu.edu/software/biofilter-download).

4. Discussion

In this study, we proposed a genetic interaction analysis framework of using human phenotypenetwork (HPN) as knowledge-based data filter and subsequently statistical epistasis network(SEN) for quantitative epistatic interaction detection and visualization. Our method was suc-cessfully applied to GLAUGEN GWAS data and we were able to identify a connected networkof interacting SNPs that included significantly more SNPs than expected randomly. Such a

rs6571228

rs5768690

rs9941233

rs953259

rs956627

rs992014rs9366399

rs10814916

rs7678436

rs2822557

rs4640244

rs9841585

rs4369779

rs13266634

rs738494

rs10777084

rs2274557

rs1320288 rs3752591

rs13201929

rs10165941

rs5219

rs17141444

rs9563035

rs2079918

rs7466269

rs9557082

rs10505031 rs813164

rs10486771 rs7138792

rs1077165

rs4653061 rs10184275

rs3752823

rs2462140

rs10010325 rs10008860

rs10777845 rs245136

rs2277142 rs2277912

rs9374274

rs1401471

rs4743034 rs16872571

rs2605463

rs9472138

rs9349200

rs1570211

rs11669012

rs17706439

rs1771428

rs189798

rs6996971

rs1504749

rs7336332

rs1396477

rs12517906

rs1220025

rs11714343

rs11664670

rs150724

rs2449222

rs569688

rs7209435

rs930716

rs11580979 rs1533971

rs2732494

rs12076129 rs5015480

rs10505879

rs10508288

rs9935794

rs1816002 rs4455882

rs17103138

rs882300

rs6663404

rs2162256

rs11782819 rs284489

rs989488

rs9487094

rs10502372

rs10509779

rs2281680

rs12199222

rs4897336 rs7172432

rs4313471

rs2273194

rs4121677

rs7542900 rs7178902

rs10237118 rs11108037

rs739981

rs551719

rs1035025

rs688391rs2621469

rs6982250

rs2106913

rs3924153

rs1491192

rs1436900

rs1553847 rs1270837

rs756199

rs7017184 rs2291256

rs597425rs11730157

rs3125050 rs1490388

rs1568482

rs1524976

rs3825214 rs1251588

rs3748069

rs7914674

rs4371861

rs972869

rs891088

rs10848958

rs7525553

rs886374

rs892267

rs888663

rs894558

rs891835

rs896076

rs3916765

rs747175

rs353121

rs3743200

rs3743157 rs1858943

rs3745060

rs7171517

rs1935881

rs7249094

rs300483rs3805647

rs4537554

rs391300

rs1245531 rs3118914

rs173700rs2844479

rs11981985

rs7191006

rs710841

rs6885224

rs2149998

rs10499480

rs12502295

rs1037124

rs7318731

rs3766793

rs17034592

rs10468542 rs4242252

rs2580816 rs11856676

rs12741973

rs9449444

rs344907

rs10506546

rs1746048

rs7128948

rs2783169

rs17084051

rs6918956

rs923375

rs4503747

rs10906115

rs11154022

rs2286036

rs2782934

rs9946041 rs6764363

rs270504

rs10791993

rs2347824

rs1367158

rs1805015 rs1802295

rs4721923

rs2853552

rs2171

rs2849460

rs10518996

rs4722613

rs943855

rs10506627 rs1239947 rs7756992

rs944260

rs540937rs8048899

rs9326439

rs2522425

rs13097792

rs538827

rs4771591

rs282544

rs2015454

rs2829588 rs2522422

rs4672150

rs1558139

rs2750023

rs6020814

rs6504218

rs553346

rs10014145

rs322239

rs7743494

rs7502462

rs2670321

rs1532357

rs2948998

rs17171705

rs10424480

rs3827414

rs1944866

rs1949009

rs300493

rs6725887

rs7403531

rs10014970

rs4365558

rs11211081

rs7025486

rs13438327 rs334213

rs1880781

rs188019

rs1040046

rs9568793 rs1901103

rs8055236

rs535343

rs4951536

rs16852525

rs2024341

rs785422

rs517871

rs11131055

rs11123849

rs13077378

rs11708189

rs2074404

rs6585827

rs10502032

rs17286397

rs1950500

rs3825569

rs1945085

rs499818

rs11170631

rs7069291

rs1058793

rs2306374 rs7812590

rs1875206

rs1874326

rs2410501

rs2627775

rs727957

rs1950702

rs11122547

rs9626074

rs4793501

rs745557rs10829156

rs17356672

rs11926571 rs12592821

rs7954465

rs655487

rs1293151

rs1438342

rs12356233 rs10518527 rs2319517

rs1410518

rs10863936

rs1406961

rs804125

rs974110

rs11807640

rs10518622

rs7873584

rs7314446

rs2383208

rs6435396

rs3774937

rs11241936

rs9391221 rs560713

rs6499640

rs1847118 rs1407149

rs248471

rs1137

rs13012046

rs12576239

rs7833986

rs4982323 rs2478830

rs12150284

rs1229542 rs7669668

rs4842838 rs498872 rs1017307

rs1029973

rs11090754

rs7498403

rs6905922

rs10801913

rs3797235

rs6476642

rs928952

rs4792192

rs6129032

rs7810842

rs2385576

rs1399004

rs212062

rs11223996

rs11127920

rs10496091

rs13389923

rs1545552

rs1275419 rs1052131

rs1865640

rs2146758

rs1733724

rs1007000

rs10922273 rs2470984

rs1111233

rs6780569 rs10895547

rs1525361 rs10085496

rs10834309

rs11599750

rs925019

rs2853676

rs2876303

rs4379368

rs745324

rs6779469

rs2753268

rs1538138

rs1333040

rs4687161

rs17565975

rs914981

rs13123790 rs12651137

rs11921014 rs7239883

rs7669856

rs6572380

rs7754614 rs706080

rs10504130

rs7129429

rs9328448

rs9323914

rs1192415

rs4507859

rs6708562

rs2750024

rs1357523

rs9897551

rs2833836 rs9317297 rs10497323

rs1467242 rs1549250

rs1444439

rs7003283

rs1741344

rs6107955 rs7823896 rs11918414

rs779388 rs6010620

rs2171963

rs10509680

rs10510189

rs297367

rs7084752

rs7481311 rs1057510 rs1540923

rs918911

rs16852600

rs7792939

rs9498099 rs766992

rs2123808

rs4800148

rs2142219 rs1039524

rs2221618

rs1151008

rs2701248 rs10518221

rs7028544

rs4978434

rs11196686 rs1434709

rs11943513

rs7090871

rs1048228 rs701276

rs1325273

rs9863717

rs798489

rs9285397

rs803422

rs7031748 rs10932055

rs6895327

rs10807323

rs1859428

rs10816943

rs9540221 rs2047870

rs1861612

rs10738886

rs10828213

rs12359685

rs2223662

rs13102884

rs13117608

rs2080401

rs7208097

rs2048327

rs10493900

rs1559777

rs12418204

rs6566726

rs8190893

rs1167242

rs26438

rs7604827

rs4421636

rs6565681 rs1473533

rs1324034 rs1012997

rs10929060 rs10485733 rs6795970

rs6699417

rs834485rs3111828

rs17550532 rs5751614

rs2367563

rs1145153

rs831571

rs1145796

rs831983

rs7183263 rs12338076

rs10520520 rs815815

rs6284

rs968008

rs4380028

rs788178rs7856675

rs705384

rs7044529

rs2236691

rs7084584

rs1965072

rs4979264 rs9521112

rs1216472

rs7029757

rs7669522 rs13240504

rs2554380 rs9399799

rs284495

rs494459 rs3134904

rs7045593 rs704744

rs1044581

rs167275

rs1105778

rs17261321 rs4845405

rs1324982 rs2439312

rs4778818 rs16940375

rs17553657 rs3027322

rs10504592 rs1635281 rs472265

rs7648727 rs1376486

rs2547416

rs1022141

rs17041869

rs964184

rs10483727 rs3129055

rs7916697 rs9644059

rs17128272

rs557962

rs6856061

rs795955rs9294851

rs6013871

rs4129959

rs38831

rs1889151

rs12597778

rs12420080

rs1260836

rs1888814

rs3786800

rs7665764

rs16861329 rs885389

rs16860271

rs4952125

rs1412444 rs1932397

rs1412829

rs1935681

rs6774494

rs6780510

rs5215

rs7606754

rs280171

rs7112925

rs6773931

rs1230545

rs295208

rs9521509

rs6869688 rs1933988

rs8111071

rs10495936

rs1076673

rs17584499

rs1267043

rs4666636 rs7013518

rs4346352

rs1541010

rs1480577

rs1719079

rs1754073 rs1490331

rs10754458 rs9359617

rs4405822

rs587847

rs9358118

rs16893890

rs9296824 rs10282118

rs6671793

rs12358475

rs1870661 rs2168159

rs16897288

rs2206734

rs6667140

rs6670655

rs11656696

rs1293470

rs9325777

rs2717128

rs919129

rs7741993 rs7488278

rs642773

rs7963889

rs1476882

rs7742369

rs7651039

rs758099

rs10491472

rs6507838 rs1363258

rs12940030

rs6570507

rs6765739

rs12679004

rs4527850

rs1435066 rs10915424

rs10430923

rs6929078

rs4977574

rs13150558

rs7092929 rs792747

rs2794467

rs8000245

rs16886072

rs9825379

rs1410811

rs1898162 rs2390024

rs11782634 rs913170

rs9386234

rs7809017

rs2405781

rs140174

rs12682841

rs11861962

rs11859036

rs668204

rs1484948

rs17364464

rs3771395 rs941567

rs1549318

rs17291650

rs979976

rs343079

rs4794665 rs9318225 rs7660418

rs3847375

rs11900940 rs6564651

rs6480314 rs6563943 rs989548

rs1572072

rs274917

rs4072683

rs2188857

rs2376999

rs7000183

rs6036896

rs13167932 rs952361rs17288067

rs1281317

rs16907259

rs2767000

rs11062784

rs1870301

rs2501968 rs10502182 rs1968011 rs1019156

rs12583288 rs7332115

rs7698623 rs12579680

rs2305707

rs1315134

rs4130328

rs12931996

rs10514431 rs9304456

rs12970134

rs565659rs1053049

rs7404645

rs2495699

rs9848261

rs4648011 rs7027110

rs756202

rs10505784

rs2839670

rs1532358

rs17117825

rs2273301

rs1998038

rs6792584 rs12446047

rs463235

rs10511904

rs10496166

rs7283906 rs11763147

rs11749320 rs10758892 rs400337

rs2869832

rs10514995

rs2967055

rs342169rs8042680

rs6981943

Fig. 3. The SEN of Glaucoma. The network includes 713 SNPs (vertices) and 789 pairwise interactions(edges). The size of a vertex represents the strength of the main effect of its corresponding SNP, with thedisease association ranging from �0.001% to 1.124%. The width of an edge indicates the strength of itscorresponding interaction, with the disease association ranging from 0.693% to 1.163%.

degree

frequency

1 2 3 4 5 6 7 8

0

50

100

150

200

250

betweenness

frequency

0.0 0.1 0.2 0.3 0.4 0.5

0

100

200

300

400

500

closeness

frequency

0.03 0.04 0.05 0.06 0.07 0.08

0

20

40

60

80

100

120

140

(a) (b) (c)

Fig. 4. The distributions of (a) degree, (b) betweenness, and (c) closeness centralities of vertices in the SENof Glaucoma.

large connected SNP network indicated a complex genetic basis of glaucoma.HPN is an computational model that systematically organizes and clusters various pheno-

types and diseases based on their shared genetic association factors or related basic functionalunits. It also serves a powerful knowledge-based filter for GWAS data pre-screening and pri-oritization. Indeed, when compared to established data-driven filters, including the family ofRelief algorithms, the HPN filtering performed equally well. This was evaluated by quantifyingall the pairwise interactions of filtered sets of SNPs using different data-driven filters (datanot shown).

SEN is able to detect and analyze genetic interactions among a large set of attributes ingenetic association studies. The derived SNP interaction network of glaucoma in this studyhad a significant connected structure and may serve as a map for identifying key geneticfactors associated with the disease with further biological validations. The significant negativemain effect assortativity of such a network indicated that epistatic interactions were likely tohappen between SNPs with negatively correlated main effects. This finding suggests that mostexisting main-effect-based GWAS data filtering algorithms may overlook potential importantinteractions since some SNPs with low main effects would not be pre-selected using such filters.

The set of SNPs with highest ranked centralities were annotated using Biofilter33 to theirassociated genes and, in turn, the genes to pathways and functional groups. We used theannotation to identify other complex diseases possibly sharing biology with glaucoma andsuggest novel hypotheses for undocumented shared genetic interactions.

It is known that glaucoma and type II diabetes (T2D) are associated, but direct geneticrelationship remains unknown. We found SNPs connecting T2D to glaucoma: rs1333040 (CD =

5, CB = 0.043) and rs1053049 (CD = 5). The genes CDKN2B-AS1 and PPARD, which these twoSNPs are respectively located in, are known to be associated with diabetes. Peng et al. observedthat CDKN2A/B gene had a 1.17-fold (95% CI: 1.1-1.23) risk estimate for T2D.34 PPARDhas been found to be associated with higher fasting glucose, glycated hemoglobin (HbA1c),and increased risk of T2D and combined impaired fasting glucose/T2D in a population-basedChinese Han sample.35 SNP rs9540221 (CD = 5) is also intergenic between NFYAP1 and

LGMNP1, of which the former has also been associated with diabetes and obesity. Indeed,population-based studies suggest that persons with T2D have an increased risk of developingopen-angle glaucoma (OAG): the Beaver Dam Eye Study,36 the Framingham Eye Study,37

and the Los Angeles Latino Eye Study.38 Chopra et al. has noted that there is a positiveassociation between T2D and OAG in a sample Latino population study in Los Angeles,USA.38 The Rotterdam Study39 has also previously reported a positive association betweenDM and OAG, with a relative risk of 3.11 (95% CI, 1.12-8.66). Laboratory evidence for T2Dand glaucoma linkages also exists. Some have observed that there is an inner retinal celldeath, determined by the presence of hyaline bodies in the ganglion cell and retinal nerve fiberlayer of the eyes of post-mortem eyes of patients with T2D.40 Possible hypotheses of T2D toglaucoma direct linkage include how altered biochemical pathways and vascular changes inT2D reduce blood flow and impaired oxygen diffusion, thus increasing oxidative stress andendothelial cell injury.41 Excessive glial cell activation may contribute to chronic inflammation,making the retinal ganglion cells more susceptible to glaucomatous damage.42 Connectivetissue remodeling might also promote increased intraocular pressure (IOP) and greater opticnerve head mechanical stress, which can result in glaucomatous optic neuropathy.43 Diabetesis known to cause microvascular damage and may affect the vascular autoregulation of theoptic nerve and retina. Thus a longer duration of T2D would be associated with a higher riskof OAG.

Obesity, like T2D, is a trait often associated with glaucoma, but whose connection remainselusive. We found associations between obesity and glaucoma in SNPs rs12970134 (CD =

5, CC = 0.062), rs11807640 (CD = 8, CB = 0.516, CC = 0.071), rs9540221 (CD = 5), rs10777845(CD = 5), and rs4365558 (CD = 5, CC = 0.063), all of which showed significance in our glaucomastudy. SNP rs12970134 is located upstream of gene MC4R, the defects of which have beenidentified as a cause for autosomal dominant obesity.44 SNP rs11807640 is located downstreamto gene CDK4PS, which has been associated with obesity traits among some postmenuposalwomen.45 Both rs9540221, which is intergenic between NFYAP1 and LGMNP1 and rs4365558,which is intergenic between A4GALT and RPL5P34, are SNPs situated between genes thathave been associated with obesity as well. Finally, rs10777845 is upstream to gene RMST,which has been observed to be associated with severe early-onset obesity.46 In the investigationto determine direct linkage between glaucoma and obesity, several hypotheses exist. Newman-Casey et al. observed that not only was the frequency of OAG was higher among obeseindividuals (3.1%) than among non-obese individuals (2.5%), but that the relationship betweenOAG and obesity had significant correlations with gender.47 Obese women had a 6% increasedhazard of developing OAG when compared to non-obese women, while there was no significanteffect in men. Similar findings were made by Zang and Wynder, in which they observed OAGdiagnosis as twice as likely in women with a body mass index (BMI) above 27.5, but no impactof BMI in men.48 Some hypothesize that increased orbital pressure as a result of excess fattissue may cause a rise in venous pressure and a consequent increase in intraocular pressure(IOP).47

Our study also show links between Alzheimer’s disease (AD) and glaucoma. The SNPrs803422 (CD = 5, CC = 0.060), which we found to affect glaucoma is also associated with

AD. In addition, the two genes SORCS3 and MTHFD1L which contain two of the SNPs withthe strongest association are also known to be associated with AD. Patients with AD havea significantly increased rate of glaucoma occurrence. In a study in four nursing homes inGermany, 112 Alzheimer’s patients were taken as a case group. 29 of those 112 were found tohave Glaucoma, a rate of 25.9% as opposed to a 5.2% rate in the control group.49 However,in spite of significant research on this topic, there have still been no clear clinical or geneticrelationships found between the two diseases.50

Perhaps unsurprisingly, our research has also led to us connecting glaucoma with otherphenotypes relating to the eye, particularly cataracts and myopia. The association betweencataracts and glaucoma can be seen in the similar genes CDK4PS , NFYAP1, and LGMNP1,all of which contain SNPs found to be highly associated with glaucoma. Myopia and glaucomaare both directly related to the SNP rs569688 (CB = 0.415, CC = 0.074). There have been manyrecent studies about the connection between glaucoma and myopia, particularly primary open-angle glaucoma (POAG) and high myopia. Though the connection has still not been found, itis believed the high myopia is another risk factor of POAG, though it is not as important asintraocular pressure. However, it is recommended that those with high myopia be screen forglaucoma more frequently.51

Of the SNPs and genes associated to glaucoma in our study, the majority that are notdirectly eye-related, are heart related. Glaucoma is connected therefore probably connectedto cardiovascular disease, coronary artery disease, and coronary restenosis. SNPs rs17034592(CD = 7, CC = 0.060) and loci LDB2, CDKN2B and TRNAQ46P are all associated withcoronary artery disease. SNPs rs499818 (CD = 5, CC = 0.052), rs1333040 (CD = 5, CC = 0.052),and loci A4GALT, RPL5P34, and CST7 are all associated with myocardial infarction, andA4GALT and RPL5P34 are associated with cardiovascular disease. Studies have shown thatcardiovascular disease may play an important role in the development and progression ofglaucoma. However, it is also hypothesized that this relationship may be indirect, due tothe variable of age.52 Additional studies about the relationship between intraocular pressure,the main risk factor for glaucoma, and cardiovascular disease exist. However, any geneticrelationship between these two diseases is yet to be found.

An undocumented interaction found in this study is the link between glaucoma and col-orectal carcinoma. Relevant SNPs that were significant were rs972869 (CD = 7, CC = 0.052),located near the EPHB1 gene, and rs300493 (CD = 6, CB = 0.311, CC = 0.066), located inthe gene NAV3, both of which are linked to colorectal carcinoma. Studies have shown thatobesity and T2D are associated risks of colorectal cancer, which could lead to the hypothesisthat colorectal carcinoma and glaucoma may be co-morbidities and are affected by similarpathways.

In summary, the successful application of our methodology on GLAUGEN GWAS dataproves the effectiveness of our bioinformatics framework, and suggests its great potential indetecting and characterizing gene-gene interactions in GWAS data. In future studies, we expectto extend the interaction analysis beyond the order of pairs. The increased computationaldemand would, however, require a stricter data pre-filtering. The more stringent SNP filteringcan be achieved using a HPN based on SNP overlap to restrict the interactions to the most

significant ones. Other avenues of research would include using eye-specific tissue genotypedata to further refine the analysis and the drug response of the nodes of interest.

Acknowledgements

This work was supported by the National Institute of Health (NIH) grants R01-EY022300,R01-LM010098, R01-LM009012, R01-AI59694, P20-GM103506, P20-GM103534.

References

1. R. Sachidanandam, D. Weissman, S. C. Schmidt, J. M. Kakol, L. D. Stein and et al., Nature409, 928 (2001).

2. W. Y. S. Wang, B. J. Barratt, D. G. Clayton and J. A. Todd, Nature Review Genetics 6, 109(2005).

3. H. J. Cordell, Human Molecular Genetics 11, 2463 (2002).4. J. H. Moore, Nature Genetics 37, 13 (2005).5. N. A. Davis, J. E. Crowe Jr., N. M. Pajewski and B. A. McKinney, Genes and Immunity 11,

630 (2010).6. T. Hu and J. H. Moore, Network modeling of statistical epistasis, in Biological Knowledge Dis-

covery Handbook: Preprocessing, Mining, and Postprocessing of Biological Data, eds. M. Elloumiand A. Y. Zomaya (Wiley, 2013) pp. 175–190.

7. A. Pandey, N. A. Davis, B. C. White, N. M. Pajewski, J. Savitz, W. C. Drevets and B. A.McKinney, Translational Psychiatry 2, p. e154 (2012).

8. J. L. Wiggs, M. A. Hauser, W. Abdrabou, R. R. Allingham, D. L. Budenz, E. Delbono, D. S.Friedman, J. H. Kang, D. Gaasterland, T. Gaasterland, R. K. Lee, P. R. Lichter, S. Loomis,Y. Liu, C. McCarty, F. A. Medeiros, S. E. Moroi, L. M. Olson, A. Realini, J. E. Richards, F. W.Rozsa, J. S. Schuman, K. Singh, J. D. Stein, D. Vollrath, R. N. Weinreb, G. Wollstein, B. L.Yaspan, S. Yoneyama, D. Zack, K. Zhang, M. Pericak-Vance, L. R. Pasquale and J. L. Haines,J Glaucoma 22, 517 (Sep 2013).

9. M. Takamoto and M. Araie, Ophthalmology Journal 58, 1 (2014).10. J. H. Fingert, Eye (Lond) 25, 587 (May 2011).11. K.-I. Goh, M. E. Cusick, D. Valle, B. Childs, M. Vidal and A.-L. Barabasi, Proceedings of the

National Academy of Sciences 104, 8685 (2007).12. C. Darabos, K. Desai, R. Cowper-Sallari, M. Giacobini, B. Graham, M. Lupien and J. Moore,

Lecture Notes in Computer Science 7833, 23 (2013).13. C. Darabos, M. J. White, B. E. Graham, D. N. Leung, S. Williams and J. H. Moore, BioData

Mining 7, p. 1 (Jan 2014).14. H. Li, Y. Lee, J. L. Chen, E. Rebman, J. Li and Y. A. Lussier, Journal of the American Medical

Informatics Association : JAMIA 19, 295 (January 2012).15. D. Welter, J. MacArthur, J. Morales, T. Burdett, P. Hall, H. Junkins, A. Klemm, P. Flicek,

T. Manolio, L. Hindorff and H. Parkinson, Nucleic Acids Res 42, D1001 (Jan 2014).16. T. Hu, N. A. Sinnott-Armstrong, J. W. Kiralis, A. S. Andrew, M. R. Karagas and J. H. Moore,

BMC Bioinformatics 12, p. 364 (2011).17. T. Hu, Y. Chen, J. W. Kiralis and J. H. Moore, Genetic Epidemiology 37, 283 (2013).18. T. M. Cover and J. A. Thomas, Elements of Information Theory: Second Edition (Wiley, 2006).19. A. Jakulin and I. Bratko, Analyzing attribute dependencies, in Proceedings of the 7th European

Conference on Principles and Practice of Knowledge Discovery in Databases (PKDD 2003), ,Lecture Notes in Artificial Intelligence Vol. 2838 (Springer-Verlag, 2003).

20. J. H. Moore, J. C. Gilbert, C.-T. Tsai, F.-T. Chiang, T. Holden, N. Barney and B. C. White,Journal of Theoretical Biology 241, 252 (2006).

21. D. Anastassiou, Molecular Systems Biology 3, p. 83 (2007).22. J. H. Moore, N. Barney, C.-T. Tsai, F.-T. Chiang, J. Gui and B. C. White, Human Heredity 63,

120 (2007).23. B. A. McKinney, J. E. Crowe, J. Guo and D. Tian, PLoS Genetics 5, p. e1000432 (2009).24. M. E. J. Newman, Networks: An Introduction (Oxford University Press, 2010).25. M. E. J. Newman, Physical Review Letters 89, p. 208701 (2002).26. L. C. Freeman, Sociometry 40, 35 (1977).27. A. Bavelas, Journal of the Acoustical Society of America 22, 725 (1950).28. G. Sabidussi, Psychometrika 31, 581 (1966).29. J. H. Moore and B. C. White, Lecture Notes in Computer Science 4447, 166 (2007).30. C. S. Greene, N. M. Penrod, J. Kiralis and J. H. Moore, BioData Min 2, p. 5 (2009).31. X. Sun, Q. Lu, S. Mukheerjee, P. K. Crane, R. Elston and M. D. Ritchie, Front Genet 5, p. 106

(2014).32. C. International HapMap, Nature 437, 1299 (Oct 2005).33. S. A. Pendergrass, A. Frase, J. Wallace, D. Wolfe, N. Katiyar, C. Moore and M. D. Ritchie,

BioData Min 6, p. 25 (2013).34. F. Peng, D. Hu, C. Gu, X. Li, Y. Li, N. Jia, S. Chu, J. Lin and W. Niu, Gene 531, 435 (Dec

2013).35. L. Lu, Y. Wu, Q. Qi, C. Liu, W. Gan, J. Zhu, H. Li and X. Lin, PLoS One 7, p. e34895 (2012).36. B. E. Klein, R. Klein and S. C. Jensen, Ophthalmology 101, 1173 (Jul 1994).37. P. Mitchell, W. Smith, T. Chey and P. R. Healey, Ophthalmology 104, 712 (Apr 1997).38. V. Chopra, R. Varma, B. A. Francis, J. Wu, M. Torres and S. P. Azen, Ophthalmology 115, 227

(Feb 2008).39. S. de Voogd, M. K. Ikram, R. C. W. Wolfs, N. M. Jansonius, J. C. M. Witteman, A. Hofman

and P. T. V. M. de Jong, Ophthalmology 113, 1827 (Oct 2006).40. J. R. Wolter, Am J Ophthalmol 51, 1123 (May 1961).41. V. H. Y. Wong, B. V. Bui and A. J. Vingrys, Clin Exp Optom 94, 4 (Jan 2011).42. M. Nakamura, A. Kanamori and A. Negi, Ophthalmologica 219, 1 (Jan-Feb 2005).43. V. C. Lima, T. S. Prata, C. G. V. De Moraes, J. Kim, W. Seiple, R. B. Rosen, J. M. Liebmann

and R. Ritch, Br J Ophthalmol 94, 64 (Jan 2010).44. K. M. Ling Tan, S. Q. D. Ooi, S. G. Ong, C. S. Kwan, R. M. E. Chan, L. K. Seng Poh, J. Mendoza,

C. K. Heng, K. Y. Loke and Y. S. Lee, J Clin Endocrinol Metab 99, E931 (May 2014).45. P. Coveney and R. Highfield, Frontiers of Complexity: The Search for Order in a Chaotic World

(Faber and Faber, London, 1995).46. E. Wheeler, N. Huang, E. G. Bochukova, J. M. Keogh, S. Lindsay, S. Garg, E. Henning, H. Black-

burn, R. J. F. Loos, N. J. Wareham, S. O’Rahilly, M. E. Hurles, I. Barroso and I. S. Farooqi,Nat Genet 45, 513 (May 2013).

47. P. A. Newman-Casey, N. Talwar, B. Nan, D. C. Musch and J. D. Stein, Ophthalmology 118,1318 (Jul 2011).

48. E. A. Zang and E. L. Wynder, Nutr Cancer 21, 247 (1994).49. A. U. Bayer, F. Ferrari and C. Erb, Eur Neurol 47, 165 (2002).50. P. Wostyn, K. Audenaert and P. P. De Deyn, Br J Ophthalmol 93, 1557 (Dec 2009).51. S.-J. Chen, P. Lu, W.-F. Zhang and J.-H. Lu, Int J Ophthalmol 5, 750 (2012).52. S. S. Hayreh, Survey of Ophthalmology 43, Supplement 1, S27 (1999).