+ - 0:00:00
Notes for current slide
Notes for next slide

The disruptive nature of portable sequencing for community empowerment… and GPUs

Genomics Aotearoa Friday Seminar Series, 9th April 2021

Dr Miles Benton

Human Genomics
Institute of Environmental Science and Research (ESR)
Aotearoa (New Zealand)

1 / 53

Who are we?

You may have seen recently...

4 / 53

Genomics in NZ's pandemic response

A sequencing success story made possible by the awesome Una Ren, Matt Storey & Joep de Ligt

Give us some time and money and we'll demonstrate what this technology can do.


5 / 53

Genomic sequencing

6 / 53

DNA - A, T, C, G Part of all 'living' things

A snapshot of the Human Genomics group

  • Independent research
    • type-2 diabetes & obesity (genomics, epigenetics, protein 'omics)
    • Guinea pigs! (Max Berry, University of Otago)
    • empowering people to explore and interpret their data


  • Clinical diagnostics through genetics
    • Contract(s) with CCDHB Wellington Regional Genetics Lab
    • Providing support to CCDHB for Clinical Genomics GA project (Stephen Robertson)


  • Clinical research, in collaboration with local clinician (Richard Carroll) - hyperparathyroidism
    • sequencing 100 exomes
    • engaging clinicians from other DHBs
    • exploring the replacement of gene panels with exome/genome
7 / 53

A big part of what we do is sequencing

8 / 53

(image source: illumina.com)

9 / 53

I used the word portable in my previous slide, this is not a phrase one would attribute to machines from Illumina / PacBio.

These things are huge! ... and expensive.

"Soon, It Will Cost Less To Sequence A Genome Than To Flush A Toilet -- And That Will Change Medicine Forever"

10 / 53

Can we speed things up?

Thanks to eResearch 2020 ...enter Nvidia Clara Parabricks

powered by GPU

11 / 53

How much does that matter?

Identical pipelines with a slight difference...

  • Pipeline [CPU]
    • Exomes: >100X exome takes between 8-12 hours
    • Genomes: ~30X genome takes between 30-48 hours

  • Pipeline [GPU - Parabricks]
    • Exomes: >100X exome takes between 6-12 mins
    • Genomes: 100X genome takes ~4 hours

    • was on 2x V100 GPUs, now also have 2x A100s
13 / 53

15 / 53

Complete Overhaul! [Leah Kemp]


ESR GitHub: https://github.com/ESR-NZ

17 / 53

Another type of sequencing technology

...sorry PacBio I'm ignoring you today...

18 / 53

(image source: nanoporetech)

19 / 53

flongle: ~$500 USD for 5 1 might be enough for a metagenome

minION: $1000

smidgION: still just a concept

gridION: ~$50K

promethION: ~$230K

David's talk and a live demo are linked.

Example squiggle plot

(image source: tombo manual)

22 / 53

What we 'see' when DNA passes through the pores.

GPUs accelerate basecalling of Nanopore long-read data (a lot)


Nanopore basecalling benchmarking: link (ESR GitHub)

23 / 53

Why GPUs?

New type of sequencing data requires new type of 'analysis'

  • squiggle data lends itself nicely to neural nets
  • GPUs are very capable in this space
    • CUDA cores

ESR GPU basecalling benchmarks (link)

  • Titan RTX & 2x Tesla V100

Titan RTX


[UPDATE:] guppy is now able to scale across multiple GPUs!

24 / 53

Let's step back a couple of years

... a personal story ...

25 / 53

26 / 53

a personal story


  • had a lumbar puncture ('9 days' old)
    • (after being born 8 weeks early)


  • suspected meningitis (bacteria/viral)?
    • told after the fact


  • >48 hours for results return
27 / 53

a personal story

... imagine a more ‘real-time’ procedure:


28 / 53

a personal story

... imagine a more ‘real-time’ procedure:


  • a sample (i.e. blood) taken,
  • DNA/RNA extracted/enriched,
  • sequencing library prepared,
  • nanopore sequencing initiated,
  • sequence starts coming in almost instantly,
  • time to first read-match detection <1 hr
28 / 53

a personal story

... imagine a more ‘real-time’ procedure:


  • a sample (i.e. blood) taken,
  • DNA/RNA extracted/enriched,
  • sequencing library prepared,
  • nanopore sequencing initiated,
  • sequence starts coming in almost instantly,
  • time to first read-match detection <1 hr


This isn't a stretch of the imagination anymore, soon to be reality

28 / 53

Stress how fast and low cost this has the potential to be.

what are we doing?

Portable ‘real-time’ sequencing for the masses?

29 / 53

what are we doing?

Portable ‘real-time’ sequencing for the masses?

  • the idea

    • low cost (< many mobile phones)
    • accessible
    • portable
    • robust (...may have "performed" a drop test...)
    • fun!
29 / 53

what are we doing?

Portable ‘real-time’ sequencing for the masses?

  • the idea

    • low cost (< many mobile phones)
    • accessible
    • portable
    • robust (...may have "performed" a drop test...)
    • fun!
  • example use cases

    • field sequencing (real-time monitoring, forensics, agriculture, waterways, ...)
    • clinical settings
    • community outreach / teaching
29 / 53

what are we doing?

Portable ‘real-time’ sequencing for the masses?

  • the idea

    • low cost (< many mobile phones)
    • accessible
    • portable
    • robust (...may have "performed" a drop test...)
    • fun!
  • example use cases

    • field sequencing (real-time monitoring, forensics, agriculture, waterways, ...)
    • clinical settings
    • community outreach / teaching

Community Science | "United in Data"

29 / 53

Our experiences with the Xavier (link)

30 / 53

Talk about the Xavier specs:

  • 8 core arm
  • 16 GB of RAM
  • 512 CUDA core and 64 tensor cores
  • 512 GB SSD NVMe

Benchmarking (link)*

* spoilers for those following along live

31 / 53

Benchmarking (link)*

* spoilers for those following along live


Most Jetson devices have the ability to be put into different power modes


Note: this table represents basecalling performed when running on a portable powerbank.

31 / 53

Why is 10W mode exciting?

We're running in 30W MAX mode for this live demo - running on a portable powerpack.

Very rough comparison of Nvidia ecosystem#

Price in USD
# a very small selection of what Nvidia offers obviously!
* this can fit in an existing Jetson Nano carrier boards

32 / 53

V100 could rip through this data in a couple of mins, but consider the cost, power, etc.

Xavier NX is exciting, module that can slot into the Nano body [hold up Nano device]

Community engagement leading to empowerment

... a few stories of the journey ...

33 / 53

So how about unlocking it for everyone?

34 / 53

The unboxing and a slight snag

35 / 53

Success

36 / 53

A global cast of many


You can follow the story here: GitHub Gist - Xavier basecalling notes

37 / 53

The future...

38 / 53

The (not so distant) future...

39 / 53

Hot off the press!


© Oxford Nanopore Technologies


Jürgen Hench, Basel, Switzerland (cancer methylation sequencing)

44 / 53

Hot off the press!






Can now control all of this from our pockets!

Available now on Android and iOS
46 / 53

Our experiences with the Jeston Xaviers

Article in the NZ Herald

  • The Genius of Genomics (link)
  • Story also on ESR website (link)

Jetson Nanopore Sequencing


Xavier AGX setup and notes

Xavier base-calling benchmarking and notes

47 / 53

Our experiences with the Jeston Xaviers

Documentation


Presentations

eResearch 2020 presentation
eResearch 2021 presentation
Nvidia Spring GPU Technology Conference 2021
48 / 53

If you're interested in GPUs and genomics...

49 / 53

acknowledgements (people to 'nerd' out with!)

Matt Storey

(ESR)

Joep de Ligt

(ESR)

Donia Macartney-Coxson

(ESR)

Georgia Bell (ESR)
Una Ren (ESR)
Leah Kemp (ESR)

Matthew Walton (ESR)
Richard Dean (ESR)
Russell Smithies (ESR)

52 / 53
Paused

Help

Keyboard shortcuts

, , Pg Up, k Go to previous slide
, , Pg Dn, Space, j Go to next slide
Home Go to first slide
End Go to last slide
Number + Return Go to specific slide
b / m / f Toggle blackout / mirrored / fullscreen mode
c Clone slideshow
p Toggle presenter mode
t Restart the presentation timer
?, h Toggle this help
oTile View: Overview of Slides
Esc Back to slideshow