LICT Human-Machine-Interface

30
BIODATA ANALYSIS & VISUALIZATION Jan Aerts Faculty of Engineering - ESAT/SCD http://saaientist.blogspot.com @jandot Tuesday 1 February 2011

Transcript of LICT Human-Machine-Interface

Page 1: LICT Human-Machine-Interface

BIODATA ANALYSIS&

VISUALIZATION Jan Aerts

Faculty of Engineering - ESAT/SCD

http://saaientist.blogspot.com@jandot

Tuesday 1 February 2011

Page 2: LICT Human-Machine-Interface

Involved in genomics research:

•chicken, cow, human genome DNA sequencing

•search for genetic variation responsible for phenotype/disease

Issues with

•filtering: finding the correct set of parameters

•pattern searching: grasping the significance and effect of the mutations

=> visual analyticsTuesday 1 February 2011

Page 3: LICT Human-Machine-Interface

A. Filtering

Investigating parameter space...

Tuesday 1 February 2011

Page 4: LICT Human-Machine-Interface

putative mutations

filter 1

filter 2

A B C

different settings for filters

filter 3

Tuesday 1 February 2011

Page 5: LICT Human-Machine-Interface

AB

C

What we find...

Tuesday 1 February 2011

Page 6: LICT Human-Machine-Interface

AB

C

What we find...

State of the art: run many filter pipelines and take intersection

Tuesday 1 February 2011

Page 7: LICT Human-Machine-Interface

AB

C

What we should have found...

Tuesday 1 February 2011

Page 8: LICT Human-Machine-Interface

parameter-space

Tuesday 1 February 2011

Page 9: LICT Human-Machine-Interface

Tuesday 1 February 2011

Page 10: LICT Human-Machine-Interface

Tuesday 1 February 2011

Page 11: LICT Human-Machine-Interface

Tuesday 1 February 2011

Page 12: LICT Human-Machine-Interface

sometimes: bypass parameter-setting

Tuesday 1 February 2011

Page 13: LICT Human-Machine-Interface

Tuesday 1 February 2011

Page 14: LICT Human-Machine-Interface

Tuesday 1 February 2011

Page 15: LICT Human-Machine-Interface

Tuesday 1 February 2011

Page 16: LICT Human-Machine-Interface

Aim: use interactive visualization of the “raw” data to:

•peep inside the black box

•get feel for the data

•get feel for how filter settings influence each other

Tuesday 1 February 2011

Page 17: LICT Human-Machine-Interface

Aim: use interactive visualization of the “raw” data to:

•peep inside the black box

•get feel for the data

•get feel for how filter settings influence each other

Eradicate disease

Tuesday 1 February 2011

Page 18: LICT Human-Machine-Interface

B. Pattern searching

Making sense of our data...

Tuesday 1 February 2011

Page 19: LICT Human-Machine-Interface

Tuesday 1 February 2011

Page 20: LICT Human-Machine-Interface

Tuesday 1 February 2011

Page 21: LICT Human-Machine-Interface

Typical example: gene networks=> can we identify patterns?

same network

Tuesday 1 February 2011

Page 22: LICT Human-Machine-Interface

How do these networks differ?

Tuesday 1 February 2011

Page 23: LICT Human-Machine-Interface

Hive Plots, taken from http://mkweb.bcgsc.ca/linnet/

Tuesday 1 February 2011

Page 24: LICT Human-Machine-Interface

Aim: help researchers make sense of complicated data:

•gene networks

•structural variation in the genome

• linked data

• ...

Tuesday 1 February 2011

Page 25: LICT Human-Machine-Interface

Aim: help researchers make sense of complicated data:

•gene networks

•structural variation in the genome

• linked data

• ...

Eradicate disease

Tuesday 1 February 2011

Page 26: LICT Human-Machine-Interface

Hurdles:

•big data (millions/billions of datapoints)

=> makes interactivity difficult

solution: indexing methods, data formats, dimensionality reduction, ...

• visual encoding

Tuesday 1 February 2011

Page 27: LICT Human-Machine-Interface

Tools

Tuesday 1 February 2011

Page 28: LICT Human-Machine-Interface

User groups:

researcher => clinician => patient

Tuesday 1 February 2011

Page 29: LICT Human-Machine-Interface

Bioinformatics

Visualization

Tuesday 1 February 2011

Page 30: LICT Human-Machine-Interface

So:

•visual analytics: visually identifying patterns in large datasets to inform on statistical analysis

•use visualization to make sense of complex data

Tuesday 1 February 2011