The TMAJ Software Project and Database:

Angelo M. De Marzo MD PhDJames Morgan BS

November 12, 2007

Introduction Many putative new disease target genes with

diagnostic, prognostic, and therapeutic applications

Validation requires many samples Quantitative RT-PCR or protein arrays have

disadvantages Genes may be expressed in multiple different cell types

In situ analyses: ideal but generally slow Tissue microarrays address some of these

problems

TMA Technology

Tissue microarray technology for high-throughput molecular profiling of cancer Kallioniemi O et.al. Human Molecular Genetics, 2001, vol. 10, No. 7

Slide from Mark A. Rubin, M.D, Dana Farber

Donor Block Sampling

Transfer to Recipient Block

Tissue Microarray Advantages

High throughput Expands tissue use Uniform reaction conditions Built-in controls Economize use of reagents Facilitates data recording and linking to clinical data

JHU Tissue Microarray – 400 cores, 0.6 mm

Tissue Microarray – 400 cores, 0.6 mm each – H&E of 4 µm section

Digital Image Acquisition

Can use conventional microscopes Record data in spreadsheet: diagnoses and

interpretations Or the data can be recorded on paper for

later entry into a spreadsheet or database Major Problem:

Easy to loose track of the x and y coordinates of given spots

Tissue Microarray Image Acquisition

ACIS II, DAKOAperio ScanScope

Need for Data Management

200 TMAs from Johns Hopkins TMA lab

What is TMAJ? TMA-J is a set of open source software

tools and backend database structure to facilitate management and analysis of tissue microarrays and associated pathology and image data

What Does TMAJ Do?

Entering pathology data Managing users and permissions Designing TMAs Viewing and scoring TMA (and other)

images online Side-by-side viewing of serial TMA images

from slides stained for different biomarkers

Publishing large numbers of TMA images and datasets on the Internet

The software applications provide a platform for:

What Does TMAJ Do?

The Database Tracts: Clinical information about patients Pathology specimens and associated data Pathology tissue blocks Tissue Microarray cores TMA Blocks TMA Slides TMA core images TMA image scoring data: manual or semi-

automated

Primary Goals of System Address security issues Remove or isolate patient identifiers Manage multiple organ systems Develop web based interface Scalable to accommodate large number of

simultaneous users Storage of large sets of images with diagnoses Data structure compatible with emerging

standards for easy data exchange CaBIG compatibility (to be defined) The tissue microarray data exchange specification:

Berman et al., (http://www.pubmedcentral.nih.gov/articlerender.fcgi?artid=165444)

Database Design

Patients, Specimens, & Blocks

The Patients, Specimens, Blocks, and Tissue Diagnosis tables all form a one-to-many relationship.

ArrayBlocks

The Schema of ArrayBlock-related tables. 

Security: Protecting Patient Information

Database stored on a secure server

Identifiable patient information in encrypted tables (Approved by the IRB)

Researchers have no access to patient identifiers

Creates virtual separate entities: “clinical database” and “research database”

DATA Tissue Microarrays 701

Specimens 29,860

Tissue Blocks 34,783

ArrayCores 90,051

ArraySlides 7532

ArrayImages 248,746*230 users, 41 Institutions, updated Nov. 4, 2008

Specimens in TMAJAnimal 1484Autopsy 102Bladder 271Breast Cancer -- Resection NOS 983Colectomy 34Colon -- Biopsy 12Colon Cancer -- Resection NOS 43Control -- Human Tissue 206Duodenum Biopsy 5Duodenum Polypectomy 22Duodenum Resection 28Head/Neck 419Kidney 311Lung 292Mastectomy 110Oophorectomy 420Other 1810Pancreas -- Autopsy 119Pancreas -- Distal 32Pancreas -- Whipple 498Pancreas Cancer -- Resection NOS 540Pancreas Cancer -- Xenograft 55Prostate Autopsy, Whole 148Prostate Needle Biopsy, Clinical 1773Prostate Needle Biopsy, Research 8Radical Cystoprostatectomy 765Radical Prostatectomy 17726Simple Prostatectomy (Open Prostatectomy) 221Skin 252Transurethral Resection of Bladder 19Transurethral Resection of Prostate 145Total 29860

Tissue Blocks in TMAJ

Animal 2696Autopsy 1198Bladder 429Breast Cancer -- Resection NOS 1140Colectomy 78Colon -- Biopsy 10Colon Cancer -- Resection NOS 57Control -- Human Tissue 793Duodenum Biopsy 6Duodenum Polypectomy 36Duodenum Resection 63Head/Neck 1144Kidney 956Lung 333Mastectomy 192Oophorectomy 583Other 2497Pancreas -- Autopsy 610Pancreas -- Distal 70Pancreas -- Whipple 919Pancreas Cancer -- Resection NOS 830Pancreas Cancer -- Xenograft 104Prostate Autopsy, Whole 1073Prostate Needle Biopsy, Clinical 519Radical Cystoprostatectomy 715Radical Prostatectomy 15554Simple Prostatectomy (Open Prostatectomy) 70Skin 287Transurethral Resection of Bladder 34Transurethral Resection of Prostate 43Total 34783

Applications – Java from Sun Microsystems

Java Web Start Software

Java Web Start software provides a browser-independent architecture for deploying Java technology-based applications to the client desktop

Each application runs on a dedicated Java Virtual Machine (JVM)

Applications & Screenshots

Specimens Application

This application allows for detailed input of data on individual specimens and donor-tissue-blocks. 

Security Options: Specimens

Users may only access specimens to which they have permission.

Admins may assign a user permission to a specimen by using the Users-Specimens tab in the Administrator application.

Images Application

Image Application: Filtering

The table shows information about every image (identified by x and y) in an ArraySlide.

Images identified as “Prostate – Carcinoma” are highlighted in red.

Images Application: Viewing 2 Stains

T. Cornish, MD PhD, J. Morgan – Image Analysis Software in Process (v 1.0)

Publishing TMA Images and Scoring Data Over the Internet

Roughly modeled after Stanford Microarray Database

Concept: Once a study is published by a journal, all

TMA diagnoses, image, scoring and non-protected clinical data can be “published” as supplemental data to the Internet for public online viewing or down loading

TMAJ Images now linked to “Proteinpedia” database

(http://humanproteinpedia.org) by Akhilesh Pandy, MD PhD.

For More Information

http://tmaj.pathology.jhmi.edu To see published images

login to tmaj as a guest and then click the Images button.

Username: guest Password: guest

Institutions Using TMAJ

Johns Hopkins University Harvard Dana Farber Cancer

Institute Cleveland Clinic University of Texas Southwestern Vanderbilt University

Dynamic Fields in TMAJ

What are Dynamic Fields, why are they important, and

how are they managed in TMAJ?

Dynamic Fields

Different organ systems will have different recorded data. For example the Gleason score is only relevant to the prostate.

Dynamic fields allow TMAJ to keep track of different data for different organ systems.

TMAJ can have dynamic fields added at any time through the GUI. Database access is not needed and the code does not need to be recompiled.

Dynamic Fields GUI

When users add a new specimen, they are prompted to choose a Specimen Type. In this case they choose the “Radical Prostatectomy” type. After the type is selected, we see fields that are common for every specimen (SurgPathNumber and Date SpecimenTaken), as well as fields that are only relevant for a Radical Prostatectomy (GleasonSum, HasSeminalVesicle). Note the dynamic fields are in italics.

The user is prompted to choose a specimen type.

Changing Meta Data

Above we see a Type called “Prostate Atrophy” with several fields such as “HistologicType” and “Prostate_Zone”. The “Prostate_Zone” has several allowed choices such as “Central Zone” and “Peripheral Zone”. These values can be added, modified, or deleted by using the buttons on right.

One Approach: A Key-Values Table A Key Values table would only have 3 fields:

A Key (such a Prostate Weight), a value, and a foreign key that links the record back to the main table (such as the Specimens table).

We did not use this approach because it does not keep track of the meta-data. Meta-Data is data that describes data, and in this case it would be the type (Prostate), the field for the type (Prostate Weight), and any allowed choices.

Dynamic Data for Specimens

•Fields common to all Specimens are stored in the Specimens Table•The SpecimenTypes, SpecimenFields, and SpecimenEnums are the Meta Data•The SpecimenTypes contains values such as “Prostate”, “Bladder”, “Kidney”, and “Lung”•The SpecimenFields lists the field names for each Specimen Type. A Prostate SpecimenType may have a Gleason Score or Prostate Weight field.•The SpecimenEnums table give a list of valid choices for each SpecimenField.

TMAJ & Frida Integration

The image analysis software package Frida has been integrated with TMAJ

Using Image Analysis in TMAJ

New Image Analysis Sessions are created for a scanned array-slide

Viewing Image Analysis Results

Image Analysis Results may be viewed side-by-side with a regular scoring session

Acknowledgements

De Marzo LabJessica Hicks BSToby Cornish MD PhD

Tissue Microarray LabMarc Halushka MD PhDHelen Fedor BSMarcella Southerland BSQizhi Zheng MDJames Morgan BSKristen Lecksell BS