ICT in the IT Future of Medicine

beeuppityAI and Robotics

Oct 19, 2013 (4 years and 23 days ago)

79 views

ICT in the IT Future of Medicine
Project

Babette
Regierer

Daniel Jameson

ITFoM

Consortium

170
Partners
from

34
countries:


21 EC Member States


Associated
Countries:

Switzerland
,
Iceland
, Israel,
Croatia
, Turkey,
Norway



Other
countries:

Australia
, New
Zealand
, Canada, USA, Libanon,
Korea, Japan

Number

of

partners

in
one

country

ITFoM

-

Project vision


Assimilation of data about individuals (‘
omic
,
health records).


Incorporate these data into mathematical
models of each individual’s “health”.


Use these models to make predictions about
the health of individuals and, if necessary,
courses of treatment best suited to them.

The Virtual Patient: Integration of
various models

Molecules

Tissues

Anatomy

Statistics

Structure of
IFoM


Medical
Platform

(Kurt
Zatloukal
)




Analytical Technologies
(Hans
Lehrach
)




Infrastructure Hard
-

and

Software


(Nora Benhabiles/Oskar Mencer)




Data Pipelines
(
Ewan

Birney)




Computational Methodologies
(Mark
Girolami
)




ICT Integration
(Hans
Westerhoff
)




Coordination

and

Management


(Hans
Lehrach
/Markus Pasterk)

Challenges for ICT

Acquisition

Integration

Processing

Utilisation

Automation

Scalability

Security

Scale


12 million new cancer cases world wide / year


To address all
you would need to sequence
and
analyse

1 cancer every 2
seconds, that’s at least
two complete sequences, at least one for the
tumour

and one somatic.

Scalability


All technology must be developed with an eye
on scalability


What is appropriate now is guaranteed not to be
in 10 years


All data formats, standards and paradigms must
be flexible and extensible

Security


ITFoM

aware that a huge amount of the data
involved in the proposal was sensitive


Proposal to develop a robust, federated security
framework and policies.


Mindful of the location of data objects


certain
objects must remain within the EU.


Identity Management to build on the experience
of a variety of partners (EUDAT, UCL, EBI, IBM).

Acquisition: Data gathered

Acquisition: Data gathered


For data generation we need to consider:


-

heterologous data
produced (molecules
,

physiology
,
patient, society…?)


-

various technologies for data generation


-

different user groups (skilled vs. naïve)


-

different data management systems


-

different professional level

Acquisition: ICT to facilitate


Easy user
-
oriented process from machine to knowledge:


-

data analysis pipelines must be easy to handle and fast

(e.g. flow

computing)


-

fast data transfer systems


-

“online” data generation in the future?


-

development of automated processes


-

standards for data formats and processes


-

Suitable data management systems, data storage (local or

distributed, security issues)


-

new database structure needed to speed up data

storage,
transfer, use? (e.g. HANA system)


-

responsibility for data
curation

-

where, when, how,
who?

Integration: Pipelines to models


Complete genomes provide the framework to
pull all biological data together such that each
piece says something about biology as a whole


Biology is too complex for any
organisation

to
have a monopoly of ideas or data


The more
organisations

provide data or analysis
separately, the harder it becomes for anyone to
make use of the results


Integration: Pipelines to models


The data being gathered must be marshaled
into something useful


Processing, Storage, Retrieval


It
must be stored


It must be annotated


It must be auditable

Integration: ICT to facilitate


Federated data warehouse with
standardised

interfaces


Includes auditing services


Must integrate with security layer


Processing pipelines feed into the warehouse


Compute tasks handled on HPC platform using
already
established middleware (EBI).


Pipelines


several, draw on existing databases
for automation of annotation where possible.


Data specific compression algorithms

Processing: Simulating models


Variety of model types


Processing: Simulating models

Processing: ICT to facilitate


New algorithms and techniques.


HPC platforms.


Protocols.


New hardware.


Once size will not fit all, but all must
communicate with each other.

Utilising
: Making use of models


Closing the loop

Utilising
: Making use of models


We need to consider:


-

different target groups


-

easy access to data/information needed


-

make them work in the field/on the bedside


-

technology must be available at low price


(e.g. computing power must be cost
-
effective

= green technology)


Utilising
: ICT to facilitate


Aim is an approach that is easy
-
to
-
handle, cost
-
efficient and running on all systems


-

automated data analysis/modeling system


-

elaborated human
-
computer interface

(visualization)


-

automated updating of the information (e.g.

by text mining in publications)


-

must be easy to plug in new systems


-

legal issues


-

results instantly

ICT Components for Genomic Medicine

Healthcare Professional

Component 4

Individual query
analysis

Component 3

Additional clinical
annotation

Component 2

Genotype and
Phenotype
relationship capture

Component 1

Human sequence data
repositories

Component 5

Electronic Health Record

Component 6

Research on Clinical data

SHIP, GPRD, LSDBs,

Research Capability
Programme (RCP)

EBI:
repositories

(petabytes of genome
sequence data)

Sanger:
sequencing
(1000 genomes, uk10K)

Reference
genome sequence

~3 gigabytes

eHR system (e.g. emis):

~10 Mb Variant file

as attachment per record

Add genomics:

Up to 60 million
variant files = 600
terabytes*

Biomedical

Informatics

Institute (BII)

BII, SMEs etc.

cloud based, secure
services

Variant file

decision

support

system

open data

Personal Data

Anonymised

Data

Summary Data

Importance of Automation


Mentioned frequently in
ITFoM
.


Pipelining and
utilising

data on this scale is
impossible if all steps are conducted manually.


This includes processing, annotation,
hypothesis generation and testing.


Text mining, machine learning


No one’s actually cracked this.


Conclusions


A virtual, or digital, patient has the potential to
revolutionise

healthcare, but it will rely completely on
the creation of a broad, probably federated, IT
infrastructure.


An infrastructure such as this is non
-
trivial.


Any project as ambitious as a virtual patient requires
vastly more expertise than any one individual can hold,
but all elements of the project must interact.


Rigorous definition of data standards, interfaces and
pipelines must be coupled with a broad view of the
topology within which they play a part.