Bioinformatics

Image credit: www.biocomicals.blogspot.com

To consult the statistician after an experiment is finished is often merely to ask him to conduct a post mortem examination. He can perhaps say what the experiment died of.

R. A. Fisher. The Indian Journal of Statistics (1933-1960), Vol. 4, No. 1 (1938), pp. 14-17.

You can't fix by analysis what was screwed up by design!

Bobby L. Jones. Department of Human Genetics, University of Pittsburgh

Researchers often want Bioinformaticians to be Biomagicians, people who can make significant results out of non-significant data, or Biomorticians, people who can bury data that disagree with the researcher's prior hypothesis.

Dan Masys. Associate Clinical Professor of Medicine, UCSD

Bioinformatics is not something you are taught, it’s a way of life

Mick Watson PhD

The best bioinformaticians I know are problem solvers – they start the day not knowing something, and they enjoy finding out (themselves) how to do it. It’s a great skill to have, but for most, it’s not even a skill – it’s a passion, it’s a way of life, it’s a thrill. It’s what these people would do at the weekend (if their families let them).

The best thing about being a statistician is that you get to play in everybody's backyard.

John Wilder Tukey.

This holds true for Bioinformatics, I've been lucky enough to work with so many different collaborators.

Collaborative tools

Collaborative authorship

R

• R is great at a base level
• ...but where it shines is its extensibility through libraries
• CRAN (The Comprehensive R Archive Network)
• Currently, the CRAN package repository features 7425 available packages

RStudio

• RStudio is a IDE (Integrated development environment)
• Takes everything R and makes it BETTER!!

R for biologists

The Bioconductor repository currently features:

• 1104 available packages covering a wide range of biological context
• there are also 893 annotation packages
  • these cover a variety of technologies, but also numerous organisms

R... it makes things pretty!

Reproducible research - what do we mean?

Twiter sphere

Most important tool for #ReproducibleResearch is the *mindset*, when starting, that the end product will be reproducible. Keith Baggerly

Here we refer to...

• Longterm data storage and sharing protocols
• Within and between lab validation
• Transparent methodology (complete methodological steps provided)
• Scientific contribution beyond publication

Big Data: what is it?

© 2015 DataScience.LA - Big Ram is Eating Our Big Data

Where is Big Data?

© 2015 KDnuggets - Where is Big Data?

Data size

What makes data big and what complicates storage (opening up possibilities for the future)

Collecting and storing data

• Data capture
• Databases
• Programmatically filtering data

Data capture: REDCap

Data storage: MongoDB

/*insert collection as key:value pairs (I.e. associative array, hash map, dictionary)*/
db.collection.insert({
	name: "Miles Benton",
	affiliation: "QUT",
	tags: [		/*includes arrays*/
		"Bioinformatics",
		"Human genetics"],
	comments: {		/*includes embedded objects*/
		notes: "Simple comment",
		date: Date()}
})
/*-----------Find records---------------*/
db.collection.find(
	{ name: {$regex: "Miles"}} , 		/*add regular expressions*/
	{"comments":1,"_id":0}).pretty() 	/*pretty adds formatting*/
/*---------Formatted output-------------*/
{
	"comments" : {
		"notes" : "Simple comment",
		"date" : "Fri Nov 13 2015 11:54:02 GMT+1100 (AEDT)"
	}
}

NoSQL encompasses a wide variety of different database technologies that were developed in response to a rise in the volume of data stored about users, objects and products, the frequency in which this data is accessed, and performance and processing needs.

www.mongodb.com

Filtering data: WES

Visualisation tools

Interactivity - extensible outlay

Shinyapps.io example

• ~6000 member pedigree

Whole Genome Sequencing

© 2016 Illumina, Inc.

108 core-pedigree members

Illumina HiSeq X10 at the Garvan Institute

Many many TB of data (David Eccles)

QUT HPC facility

Total variants:

Functional* variants:

*Functional = predicted to be damaging in 5 in-silico tests (SIFT, POLYPHEN2, MUTATIONTASTER, PROVEAN, MUTATION ASSESSOR)

Founder Effect Variants

• ACOT4 (Acyl-CoA) Thioesterase 4
• Regulation of lipid metabolism
• Amino acid = p.TYR190ASN
• Freq A = 0.0099 (EUR), 0 (AFR) and 0 (EAS)
• Freq A in NI = 0.26
• Rare allele most likely came from England via Bounty Mutineers and then increased in frequency in NI due to founder effect

Founder Effect Variants

Glaucoma

There is a higher than usual prevalence of Glaucoma on NI, we explored the WGS data for assocated variants

Variant Filtering Strategy

1. filtered vcf files at 14% MAF
2. filtered based on genotypic models, both dominant (DOM) and allelic (ALL)
3. this resulted in 1083 for the DOM model and 1437 for the ALL model
4. the final filter was aimed to assess functional impact, we chose to go with SIFT and PROVEAN
5. this returned 1 variant

Variant information

• reference allele = C
• alternate allele = T
• MAF CASES = 0.43 vs MAF CONTROLS = 0.05
• ALL p-value: CC_ALL=2.849^-07

Predicted as deleterious/damaging

Gene has been previously associated with Retinoblastoma and Wilson's disease (cooper deposits around the cornea)

Diagnostics: Whole Exome Sequencing

Our lab just recieved NATA (National Association of Testing Authorities) accreditation for WES

First and only such facility in Australia offering accredited diagnosis

Ion Proton

• 2 Human Exomes
• 60-80M filtered reads
• up to 200bp reads
• 2-4 hours run

© 2016 Thermo Fisher Scientific Inc.

After QC, alignment and variant calling get between 28-42K variants per exome

Diagnostics used to take ~4 weeks from generating the data to filtering down the variants and writing a report for the clinician

We needed to make this process faster...

Custom pipeline...

• Sequencing, QC and alignment
• annotation (VEP, SnpSift, dbNSFP, dbSNP)
• variant filtering (bash and R)
• report generation (RMarkdown)

Real-time example: Exome reporting

'Beautiful' reports just a click away...