. . . . . .

.

......

Computational analysis of the ENCODE datasetsand other related epigenetic explorations

Ved Topkar

Harvard College class of 2016

Gunawardena LabHarvard Medical School

Department of Systems Biology13 August 2013

Ved Topkar (Harvard College) Epigenetic analysis at promoters 13 August 2013 1 / 1

. . . . . .

Introduction

Presentation Goals

FULL understanding of discussed materialAsk questions along the way!

Ved Topkar (Harvard College) Epigenetic analysis at promoters 13 August 2013 2 / 1

. . . . . .

Introduction

Outline

...1 Molecular biology in a jiffy

...2 A case study

Hypothesis formulationAnalyzing data

...3 More examples

Ved Topkar (Harvard College) Epigenetic analysis at promoters 13 August 2013 3 / 1

. . . . . .

Introduction

Outline

...1 Molecular biology in a jiffy

...2 A case study

Hypothesis formulationAnalyzing data

...3 More examples

Ved Topkar (Harvard College) Epigenetic analysis at promoters 13 August 2013 3 / 1

. . . . . .

Introduction

Outline

...1 Molecular biology in a jiffy

...2 A case study

Hypothesis formulationAnalyzing data

...3 More examples

Ved Topkar (Harvard College) Epigenetic analysis at promoters 13 August 2013 3 / 1

. . . . . .

Molecular Biology Essentials

The Cell

Ved Topkar (Harvard College) Epigenetic analysis at promoters 13 August 2013 4 / 1

. . . . . .

Molecular Biology Essentials

The Central Dogma

Ved Topkar (Harvard College) Epigenetic analysis at promoters 13 August 2013 5 / 1

. . . . . .

Molecular Biology Essentials

Transcriptional Regulation

Ved Topkar (Harvard College) Epigenetic analysis at promoters 13 August 2013 6 / 1

. . . . . .

Molecular Biology Essentials

Transcriptional Access

Ved Topkar (Harvard College) Epigenetic analysis at promoters 13 August 2013 7 / 1

. . . . . .

Background

Epigenetics and Gene Expression

Things beyond just the base pairs in DNA matter → gene expression

Ved Topkar (Harvard College) Epigenetic analysis at promoters 13 August 2013 8 / 1

. . . . . .

Background

The Question

Analyze the ENCODE dataset

Ved Topkar (Harvard College) Epigenetic analysis at promoters 13 August 2013 9 / 1

. . . . . .

Background

The Question

Analyze the ENCODE dataset

Ved Topkar (Harvard College) Epigenetic analysis at promoters 13 August 2013 9 / 1

. . . . . .

Background

Ved Topkar (Harvard College) Epigenetic analysis at promoters 13 August 2013 10 / 1

. . . . . .

Background

ENCODE (Overview)

.Overview..

......

National Human GenomeInstitute: Encyclopedia ofDNA Elements (ENCODE)

Nearly 600 collaboratinglabs post HGP

Ved Topkar (Harvard College) Epigenetic analysis at promoters 13 August 2013 11 / 1

. . . . . .

Background

The Data Set

Ved Topkar (Harvard College) Epigenetic analysis at promoters 13 August 2013 12 / 1

. . . . . .

Background

Data Types

Raw signals

Raw signal peak calling outputs (e.g. PeakSeq results)

Relatively course-grain peak data

Ved Topkar (Harvard College) Epigenetic analysis at promoters 13 August 2013 13 / 1

. . . . . .

A simple example TFBS

The Game Plan

.Can we reduce transcription factor binding landscapes into categories?..

......

Scan across genome, looking for promoters

Bin promoters appropriately

Score binding at each promoter

Clustering analysis

Ved Topkar (Harvard College) Epigenetic analysis at promoters 13 August 2013 14 / 1

. . . . . .

A simple example TFBS

RefSeq

.Overview..

......

Curated database of genes

New versions released asfrequently as Firefox

Includes pseudogenes,haplotype variations, andpredicted genes

Ved Topkar (Harvard College) Epigenetic analysis at promoters 13 August 2013 15 / 1

. . . . . .

A simple example TFBS

Defining Promoter?

Only upstream from TSS?

Incredibly far regulatory regions?

Intronic regulation?

Post termination regulatory elements?

1000 bp upstream

Ved Topkar (Harvard College) Epigenetic analysis at promoters 13 August 2013 16 / 1

. . . . . .

A simple example TFBS

Defining Promoter?

Only upstream from TSS?

Incredibly far regulatory regions?

Intronic regulation?

Post termination regulatory elements?

1000 bp upstream

Ved Topkar (Harvard College) Epigenetic analysis at promoters 13 August 2013 16 / 1

. . . . . .

A simple example TFBS

Defining Promoter?

Only upstream from TSS?

Incredibly far regulatory regions?

Intronic regulation?

Post termination regulatory elements?

1000 bp upstream

Ved Topkar (Harvard College) Epigenetic analysis at promoters 13 August 2013 16 / 1

. . . . . .

A simple example TFBS

Defining Promoter?

Only upstream from TSS?

Incredibly far regulatory regions?

Intronic regulation?

Post termination regulatory elements?

1000 bp upstream

Ved Topkar (Harvard College) Epigenetic analysis at promoters 13 August 2013 16 / 1

. . . . . .

A simple example TFBS

Defining Promoter?

Only upstream from TSS?

Incredibly far regulatory regions?

Intronic regulation?

Post termination regulatory elements?

1000 bp upstream

Ved Topkar (Harvard College) Epigenetic analysis at promoters 13 August 2013 16 / 1

. . . . . .

A simple example TFBS

Binning

How do we quantitatively analyze promoter presence?

Break promoter regions into bins for a finer metric?

Do we give weights to bins as a function of their position?

Single, unweighted 1000 bp bin of counts

Ved Topkar (Harvard College) Epigenetic analysis at promoters 13 August 2013 17 / 1

. . . . . .

A simple example TFBS

Binning

How do we quantitatively analyze promoter presence?

Break promoter regions into bins for a finer metric?

Do we give weights to bins as a function of their position?

Single, unweighted 1000 bp bin of counts

Ved Topkar (Harvard College) Epigenetic analysis at promoters 13 August 2013 17 / 1

. . . . . .

A simple example TFBS

Binning

How do we quantitatively analyze promoter presence?

Break promoter regions into bins for a finer metric?

Do we give weights to bins as a function of their position?

Single, unweighted 1000 bp bin of counts

Ved Topkar (Harvard College) Epigenetic analysis at promoters 13 August 2013 17 / 1

. . . . . .

A simple example TFBS

Binning

How do we quantitatively analyze promoter presence?

Break promoter regions into bins for a finer metric?

Do we give weights to bins as a function of their position?

Single, unweighted 1000 bp bin of counts

Ved Topkar (Harvard College) Epigenetic analysis at promoters 13 August 2013 17 / 1

. . . . . .

A simple example Survey of Promoters

Forward vs. Backward Strand

Ved Topkar (Harvard College) Epigenetic analysis at promoters 13 August 2013 18 / 1

. . . . . .

A simple example Survey of Promoters

TF Binding Frequency

Ved Topkar (Harvard College) Epigenetic analysis at promoters 13 August 2013 19 / 1

. . . . . .

A simple example Survey of Promoters

Histogram of promoter binding

Ved Topkar (Harvard College) Epigenetic analysis at promoters 13 August 2013 20 / 1

. . . . . .

A simple example Survey of Promoters

Promoter/TFBS intersections

Ved Topkar (Harvard College) Epigenetic analysis at promoters 13 August 2013 21 / 1

. . . . . .

A simple example Survey of Promoters

Computational Efficiency

This was an exercise in program optimization

Original algorithm took about 5 days, optimized/parallelizedalgorithm took just a few hours

Ved Topkar (Harvard College) Epigenetic analysis at promoters 13 August 2013 22 / 1

. . . . . .

A simple example Survey of Promoters

Clustering

The unsupervised grouping of information such that groups have similarelements that are dissimilar from elements in other groups

Ved Topkar (Harvard College) Epigenetic analysis at promoters 13 August 2013 23 / 1

. . . . . .

A simple example Clustering analysis

Clustering

.Minkowski Distance (p = 2 → Euclidean Distance)..

......

(n∑

i=1

|xi − yi |p) 1

p

Ved Topkar (Harvard College) Epigenetic analysis at promoters 13 August 2013 24 / 1

. . . . . .

A simple example Clustering analysis

Clustering

.Simple Linkage Clustering..

...... D(X ,Y ) = min(d(x , y)) ∀ (x , y)ϵ(X ,Y )

Ved Topkar (Harvard College) Epigenetic analysis at promoters 13 August 2013 25 / 1

. . . . . .

A simple example Clustering analysis

Clustering (simple linkage exhibiting chaining)

Ved Topkar (Harvard College) Epigenetic analysis at promoters 13 August 2013 26 / 1

. . . . . .

A simple example Clustering analysis

Clustering

.Complete Linkage Clustering..

...... D(X ,Y ) = max(d(x , y)) ∀ (x , y)ϵ(X ,Y )

Ved Topkar (Harvard College) Epigenetic analysis at promoters 13 August 2013 27 / 1

. . . . . .

A simple example Clustering analysis

Clustering (complete linkage)

Ved Topkar (Harvard College) Epigenetic analysis at promoters 13 August 2013 28 / 1

. . . . . .

A simple example Clustering analysis

Computational Efficiency

Pairwise distance calculations require a LOT of RAM

Ved Topkar (Harvard College) Epigenetic analysis at promoters 13 August 2013 29 / 1

. . . . . .

A simple example Clustering analysis

2-way clustering heatmap

Ved Topkar (Harvard College) Epigenetic analysis at promoters 13 August 2013 30 / 1

. . . . . .

A simple example Clustering analysis

2-way clustering heatmap (with random data arrays)

Ved Topkar (Harvard College) Epigenetic analysis at promoters 13 August 2013 31 / 1

. . . . . .

A simple example Roadmap Dataset Analysis

Histone Modifications (Roadmap Dataset)

Ved Topkar (Harvard College) Epigenetic analysis at promoters 13 August 2013 32 / 1

. . . . . .

A simple example Roadmap Dataset Analysis

A quick correlation test

R = 0.042

Ved Topkar (Harvard College) Epigenetic analysis at promoters 13 August 2013 33 / 1

. . . . . .

A simple example Roadmap Dataset Analysis

Conclusions

There are numerous methods of promoter binning that prove usefulfor complexity reduction

Unsupervised clustering of preprocessed promoter data yield resultswith biological significance

Ved Topkar (Harvard College) Epigenetic analysis at promoters 13 August 2013 34 / 1

. . . . . .

A simple example Roadmap Dataset Analysis

Next Steps

Incorporation of nucleosome enrichment, methylation, and histonemodification data in a more meaningful way

Further refining of cluster analysis pipeline to uncover more unknownbiology

Ved Topkar (Harvard College) Epigenetic analysis at promoters 13 August 2013 35 / 1

. . . . . .

A simple example End

Thanks!

Jeremy Gunawardena and the HMS Department of Systems Biology!

My collaborator Tobias Ahsendorf!

PRISE and Greg Llacer!

The lovely PRISE staff!

This audience!

Ved Topkar (Harvard College) Epigenetic analysis at promoters 13 August 2013 36 / 1