2. Increasing demand for FFPE for molecular characterization Genomic tumor profiling of retrospective cases growing Recent studies show tumor DNA extracted from FFPE comparable to FF for NGS FFPE archives provide high volume, longitudinal, population-based, opportunity to study ‘discontinued tumors’, can be de-identified and anonymized

3. FFPE typically associated with Surgical Pathology Reports but not structured data or annotation Locked away in Laboratory information System Unaltered text is Protected Health Information Very hard to find the right cases Difficult to accrue sufficient numbers of rare tumors and/or longitudinal cases showing specific behaviors

4. An NLP pipeline for de-identifying, annotating and storing clinical documents A system for indexing research resources (FFPE, FF, images) with document annotations An GUI for querying large repository of annotated documents and obtaining resources locally, using an honest broker model A platform to support data and tissue sharing among networks of cancer centers and other institutions

5. Approximately ten years of work, many deployed versions at multiple institutions Open source software project with contributions from others across country One year CTSA supplement supported development of Pitt-Penn Network Agreement

6. Rebecca Crowley Faculty, DBMI and UPCI Girish Chavan UI Design and Product Manager Kevin Mitchell NLP Developer Eugene Tseytlin NLP Developer Elizabeth Legowski User engagement and evaluation Michael Becich Faculty and Chair, DBMI and UPCI Roger Day, Faculty, DBMI and UPCI Carmelo Gaudioso Faculty, Oncology Director of Informatics at RPCI Carl Morrison Faculty, Pathology Director of Biobank Michael Feldman Faculty, Pathology Director of Biobank at Abramson Roni Bollag Faculty Director of Tumor Bank at GRU Abramson Tech/Tissue Teams Tara McSherry Nate DiGrigorio Joellen Weaver RPCI Tech/Tissue Teams Chris Darlak Monica Murphy Mayur Sakthivel Ken Quinn GRU Tech/Tissue Teams Colleen Cain Nancy Willis Isaac Green

7. Terminology NOBLECODER CODEABLE CHUNK FINDER SYNOPTIC DETECTOR TOKENIZERRESETTER ConTEXT ANNOTATION TRANSFORMER Cleans document, deletes existing annotations Tokenizes words, punctuation, numbers and spaces Finds synoptic text and eliminates from further processing Partition section text into sentences and phrases using ConTEXT stopwords Detects Negation, Temporality, Degree of Certainty Organizes essential output annotation in an easily processed manner NER NobleCoder optimized terminology format

13. A new, more flexible coding pipeline utilizing NOBLE coder with excellent accuracy and speed. Privacy and authorization model suitable for national network. Detailed auditing of all search and document access activity Highly scalable Java messaging based coding and indexing infrastructure to process 10s of millions of documents. Support for adding new document types and processing both pre-sectioned reports or full report text during import. User Interface improvements for researchers and administrators. Currently in beta and deployed at UPCI and Abramson. Public Release anticipated February 2014, will include VM

14. Specific Aim 1. Enhance the informatics technology to support inter-institutional “trust”, paraffin registry development, tissue microarray (TMA) development, and nondestructive tissue use Specific Aim 2. Establish the TIES Cancer Research Network (TCRN) with four founding member institutions. Develop governance, network agreements, and policies for operating the TCRN Specific Aim 3. Recruit and support pilot scientific collaborations across the network, especially focused on personalized medicine – Skin SPORES, Ovarian SPORE, RPCI-GHSU collaboration, Penn-Pitt image biomarkers Specific Aim 4. Disseminate the software and measure its impact.

15. Specific Aim 1. Enhance the informatics technology to support inter-institutional “trust”, paraffin registry development, tissue microarray (TMA) development, and nondestructive tissue use Specific Aim 2. Establish the TIES Cancer Research Network (TCRN) with four founding member institutions. Develop governance, network agreements, and policies for operating the TCRN Specific Aim 3. Recruit and support pilot scientific collaborations across the network, especially focused on personalized medicine – Skin SPORES, Ovarian SPORE, RPCI-GHSU collaboration, Penn-Pitt image biomarkers Specific Aim 4. Disseminate the software and measure its impact

16. U Pittsburgh Cancer Institute –4M SPRs available –Supports ~270 research users –Integrated with IRB certified Honest Broker system –Deployed system for Pathology (FFPE and FF) and Radiology (images) Abramson Cancer Center (Penn) –450K SPRs available –Live feed from clinical systems –Initial stages of deploying to users Roswell Park Cancer Institute –Completed data import mechanism –RPCI currently installing software –170K SPRs Georgia Regents Cancer Center –Just getting started –160K SPRs

17. All site IRB protocols now approved Single U24 Network Agreement among all sites. Penn executed. GRU and RPCI pending Authorized User Agreement to be the same at all institutions – the only regulatory step required by individual investigators to access deidentified data Agreement to use the UBMTA in unaltered form – will streamline collaborations between sites that use tissues

20. Specific Aim 1. Enhance the informatics technology to support inter-institutional “trust”, paraffin registry development, tissue microarray (TMA) development, and nondestructive tissue use Specific Aim 2. Establish the TIES Cancer Research Network (TCRN) with four founding member institutions. Develop governance, network agreements, and policies for operating the TCRN Specific Aim 3. Recruit and support pilot scientific collaborations across the network, especially focused on personalized medicine – Skin SPORES, Ovarian SPORE, RPCI-GHSU collaboration, Penn-Pitt image biomarkers Specific Aim 4. Disseminate the software and measure its impact

21. Ovarian Cancer SPORE program Study: Identify biomarkers for early stage and low malignant potential (LMP) ovarian cancer. Investigators: Dr. Kunle Odunsi (RPCI) and Dr. Bob Edwards ((UPCI) Early stage ovarian cancer has an 80-90% 5 year survival, however they comprise only ~ 20% of all cases. Currently no screening tests for early disease detection and few published attempts to identify molecular markers for early stage disease. Previous efforts to obtain FFPE by Dr. Edwards and colleagues have not yielded sufficient cases for molecular characterization. Using TIES across the two SPORE institutions, their goal is to obtain a much larger cohort for exome sequencing (Ion Torrent), as well as CNV and gene expression profiling (NanoString) of primary and recurrent tumors.

22. Representative from each of the four collaborating cancer centers Biweekly teleconferences, yearly F2F Will select and prioritize pilot projects Policies and processes for TCRN Fuel engagement and dissemination Provide institutional feedback on development roadmap

23. Specific Aim 1. Enhance the informatics technology to support inter-institutional “trust”, paraffin registry development, tissue microarray (TMA) development, and nondestructive tissue use Specific Aim 2. Establish the TIES Cancer Research Network (TCRN) with four founding member institutions. Develop governance, network agreements, and policies for operating the TCRN Specific Aim 3. Recruit and support pilot scientific collaborations across the network, especially focused on personalized medicine – Skin SPORES, Ovarian SPORE, RPCI-GHSU collaboration, Penn-Pitt image biomarkers Specific Aim 4. Disseminate the software and measure its impact

24. All 4 TIES nodes up, loaded, and operating locally and within the TCRN Steering Committee meeting regularly to develop governance policies Complete remaining software updates for operating the network First pilot projects selected and underway Requirements for block annotation Requirements for Paraffin Archive Start development on TIES v 5.2

25. TIES Team at DBMI/UPCI: Kevin Mitchell, Girish Chavan, Eugene Tseytlin, Liz Legowski Collaborators: Michael Becich (UPCI), Roger Day (UPCI), Michael Feldman (Penn), Carl Morrison (RPCI), Carmelo Gaudioso (RPCI), Samir Khleif (GRU), Roni Bollag (GRU) Collaborating technical teams and tissue bank teams Current Funding: –Advanced Development of TIES – Enhancing Access to Cancer Tissue U24 CA180921 (NCI) Previous Funding –Continued Development of TIES R01 CA132672 (NCI) –Clinical and Translational Science Institute U54 Supplement (NCATS) – Reis (PI) –Clinical and Translational Science Institute U54 RR023506 (NCATS) – Reis (PI) –Cancer Biomedical Informatics Grid (NCI) –Shared Pathology Informatics Network (NCI)

26. Additional Slides if needed

27. Identified and de-identified data stored in separate databases and potentially on separate servers. No direct access to databases other than through secure encrypted web services. Identified data web services hosted on separate Web server behind the institution’s firewall. Uses RSA 1024 cipher strength message level security for encrypting all communications between the servers and the clients. Authentication information stored on Institution’s servers, authorization information stored on common Hub server hosted at University of Pittsburgh. Role based access ensures only required level of data granularity is made available to users. Quarantine functionality built into the software to quickly and easily hide reports from all researchers. All access to the system is in the context of a protocol. All operations are logged with the activity type, user id and protocol information.

28. Each individual user will be required to meet the requirements of their organization for IRB approval. For University of Pittsburgh researchers using the current version (TIES v5) at University of Pittsburgh these include: –Preliminary Data User – sees aggregate data only “preparatory to research.” Requires sign off of HSTB. No IRB protocol. –Research Data User – sees record level deidntified data. Covered by HSTB IRB protocol. Requests are reviewed by HSTB. –Tissue User – Retrospective tissue requires IRB exempt protocol to individual investigator. Agreement to provide data and or tissue occurs on a case-by-case basis and is decided by the HSTB at the institution. Thus, the HSTB signs on to individual external protocols after reviewing all the relevant data (including IRB protocol, etc). Users of the system agree (on each login) and are trained to “quarantine” any data where stray identifiers are observed. When a report is quarantined it is no longer visible to any users until it has been “released” by the TIES development group. QA Officer assigned at each institution to review quarantined reports and correct any deidentification issues.

29. We use a commercial de-identification system called DeID TM. Pitt IRB considers use of DeID to mean sufficiently de-identified to meet Safe Harbor Institutions that do not want to use DeID can easily use other de-identifiers with TIES HIPAA notion of Neutral Third Party “Honest Broker” implemented as role in TIES TIES Honest Broker has access to all identifiers We can share training materials, sample policies and processes for UPMC TIES Honest Brokers

30. MySQL v5.6 Globus Toolkit v4 OGSADAI Web Services Pitt InCommon CA Signed Java Swing Client Java Webstart TIES Command Processors implemented as OGSADAI services handle all client requests. GTK provides secure RSA 1024 message level encryption for all communications. Separate databases host identified and de-identified data and authentication information for users. TIES Client is signed with Pitt’s InCommon Certificate, and delivered through Java Webstart