Angel Janevski - PAPAyA

14
A. Janevski, N. Banerjee, S. Kamalakaran, V. Varadan, N. Dimitrova Philips Research North America March 8, 2011 PAPAyA: Applications in Oncology Decision Support

description

PAPAyA: Applications in Oncology Decision Support

Transcript of Angel Janevski - PAPAyA

Page 1: Angel Janevski - PAPAyA

A. Janevski, N. Banerjee, S. Kamalakaran, V. Varadan, N. DimitrovaPhilips Research North AmericaMarch 8, 2011

PAPAyA: Applications in Oncology Decision Support

Page 2: Angel Janevski - PAPAyA

Philips Research North America, Angel Janevski, March 8, 2011, AMIA TBI 2011

Clinicalprognostic

indices

Tumor Biology

Drug related

toxicities

Co-morbidity

Patient circumstance

Dx & Therapy

Regiment decisions

2

Augmenting Oncology!

Personalized Tx: the keyto cancer management

Sequencing to bring personalization tests

Most favorable clinical

outcome

Page 3: Angel Janevski - PAPAyA

Philips Research North America, Angel Janevski, March 8, 2011, AMIA TBI 2011

Motivation

• Capture clinical and (new) molecular data in an integrative fashion for use by the clinician

• Present clinically actionable information in the context of the patient and relevant data (patient populations, knowledge)

• Input from clinicians: • Need for easy and meaningful access to existing data such as

clinical trials, published literature• Tools to access, quantify and visualize new data types: molecular

profiling, sequencing, underlying biology• Clinicians at institutions of different profiles have different needs• Tools to communicate reasoning to the patient

3

Page 4: Angel Janevski - PAPAyA

Philips Research North America, Angel Janevski, March 8, 2011, AMIA TBI 2011 4

PAPAyA for the Clinic

• PlAtform for Personalized Analytics Applications (PAPAyA)• Research/rapid-prototyping software framework enabling context-

sensitive access to bioinformatics and other clinical tools• As a system, PAPAyA comprises definition of the flow of the application,

the user interface, and context-specific tools to explore• We aim to provide:

i. access to and presentation of genomic profilesii. effective and seamless combination of clinical and molecular dataiii. unified and effective user interfaceiv. easy design, prototype and testing environmentv. combination of multiple tool implementation platforms to interact with

data and showcase decision support tools to clinicians

Page 5: Angel Janevski - PAPAyA

Philips Research North America, Angel Janevski, March 8, 2011, AMIA TBI 2011 5

Design of PAPAyA

• Tool interface enables flexible integration of research code running on a variety of platforms: R, Matlab, Python, Perl, Web

• Metadata-driven tool specifications and flow control• Uniform and consistent user interface• Principal Architecture:

USER INTERFACE

PRESENTATION

FLOW CONTROL

INTERNAL DATABASES

EXTERNAL DATACLINICAL KOWLEDGE

ACTION

TOOL EXECUTION ENGINE

TOOLSANNOTATION

ANALYSIS RESULTS

CONFIG AND

CONTROL

TOOL CTRL

Page 6: Angel Janevski - PAPAyA

Philips Research North America, Angel Janevski, March 8, 2011, AMIA TBI 2011 6

Units of User Interaction

• State: current status of all system components

• Context: current patient data, active tools, and accessed data

• Tool: smallest unit of substantial user action – metadata-defined external plug-ins

• Tool controls: metadata-defined possible iterations with a tool

State S/Context C

Execute Tool T

start

User changes context or state

Tool controls

Tool execution changes context

or state

Page 7: Angel Janevski - PAPAyA

Philips Research North America, Angel Janevski, March 8, 2011, AMIA TBI 2011 7

Executing and Interacting With Tools in PAPAyA

Available tools and tool groups

(dynamically updated based on current context)

Active tool

Tool control and interaction

Page 8: Angel Janevski - PAPAyA

Philips Research North America, Angel Janevski, March 8, 2011, AMIA TBI 2011 8

Tool Instance: Breast Cancer Subtype

• Extension of traditional histopathology based on disease origin cells

• Breast cancer Subtypes have distinct gene expression signatures +Notable survival differences basedon breast cancer subtypes

• Breast cancer subtypes utilization has traction with clinicians and is already part of some molecular profiling products on the market

• Still, very little indication of the kinds of tools will be needed in the clinic to contemplate data from such patient stratification

Sorlie et al 2003

Page 9: Angel Janevski - PAPAyA

Philips Research North America, Angel Janevski, March 8, 2011, AMIA TBI 2011 9

Tool Instance: Breast Cancer Subtype

• Multiple presentations of breast cancer subtype classification:1. Basic subtype assignment2. Comparative presentation of subtype calls3. Relative patient subtype “score” in a broader population4. And beyond …

Page 10: Angel Janevski - PAPAyA

Philips Research North America, Angel Janevski, March 8, 2011, AMIA TBI 2011 10

Tool Instance: Breast Cancer Subtype

• Link molecular profile to underlying biology (pathway deregulation with links to literature, targets for therapy, etc)

• Link molecular profile to epidemiological/population data

Page 11: Angel Janevski - PAPAyA

Philips Research North America, Angel Janevski, March 8, 2011, AMIA TBI 2011 11

Tool Instance: Clinical Trials

• Tools to integrate the context of the current patient with repositories– Identification of open clinical trials of relevance to current patient.

– Extract the latest clinical trial results of relevance to the current patient to help in decision-making

• Clinicaltrials.gov: open and closed studies with (somewhat) structured records with description, recruitment criteria, sites, funding, and additionally study results

• Multiple presentations of repository data:– Direct link to the web view of clinicaltrials.gov driven by PAPAyA

– Customized interpretation/visualization of the data from the repository –map of relevant clinical trials

– ..

Page 12: Angel Janevski - PAPAyA

Philips Research North America, Angel Janevski, March 8, 2011, AMIA TBI 2011 12

Tool Instance: Clinical Trials

Page 13: Angel Janevski - PAPAyA

Philips Research North America, Angel Janevski, March 8, 2011, AMIA TBI 2011

Conclusions

• Need for tools in the clinic to manage introduction of molecular profiling data particularly in the context of clinically-relevant data

• We presented our software framework PAPAyA:– Test variety of analytics workflows for breast cancer– Prototype novel clinical tools for breast cancer diagnosis and management– Two example tools for access and presentation of clinical and molecular

data: breast cancer subtyping and access to clinical trials information• We continue to develop PAPAyA and new tools for clinical workflow

integration• Future: We anticipate PAPAyA and similar solutions will help clinicians

manage new types data overload, bridge gaps between expert knowledge and common practice and make efficacious decisions

13

Page 14: Angel Janevski - PAPAyA