clinical data eav

Introduction Relational Database Clinical Annotation data CHCDB CHCDBWEB Conclusions

CHCDB & CHCDBWEB

Clinical Annotation Database and web interface

Thomas Burguiere

INSERM Unite 674

May 5th, 2011

Thomas Burguiere (INSERM Unite 674) CHCDB & CHCDBWEB May 5th, 2011 1 / 35


1 Introduction

2 Relational DatabasePrincipleBenefitsInterface

3 Clinical Annotation dataSpecificitiesE.A.V.

4 CHCDB

5 CHCDBWEB

6 Conclusions



• 1500 liver tumor samples

• Malignant (HCC) and benign (HCA) tumors

• Normal Tissue

Existing Data

• Clinical Annotations of malignant tumors (4D)• Excel files which contains :

• Clinical Annotations of malignant & benign tumors• Other annotations (mutations, clinical studies, etc.)• Tissue extractions listings (concentrations / quantities)



Existing Data

• Clinical Annotations of malignant tumors (4D)• Excel files which contains :

• Clinical Annotations of malignant & benign tumors• Other annotations (mutations, clinical studies, etc.)• Tissue extractions listings (concentrations / quantities)

Problems

• Clinical Annotations of malignant tumours can only be accessed onsingle machine

• Redundant data among di↵erent files

• Duplicated files on di↵erent machines

,! Discrepancies between di↵erent files

,! Cross-checking data between the di↵erent data source is cumbersome



1 Introduction



4 CHCDB

5 CHCDBWEB

6 Conclusions



Principle

Relational Database : Software

• Relational Database Management System : software which containsand organizes data (OracleTM, MySQL, DB2TM, SQL ServerTM, etc.)

• Client Server architecture :• Server software, which manages data, installed on a single machine• Client software, which queries the server, installed on any machine used to

consult the database



Principle

Client Server architecture

��

��

��



Principle

Relational Database : Data

• The data is stored in a set of tables

• One can define a set of constraints regarding the data contained inthe tables

• The tables can be associated to one another by logical links : integrityconstraints



Principle

Relational Database : Example

!"##$%&' !$()*!+,% -%. /0% -1%21)#"# 34526%3)(2# 789

!"!#$%& "!' ( )# &*+, - ./.

!"!)01& "!! ( 2-

!"!)#.& "!! 3 $.

Classical table (e.g. Excel sheet)



Principle

Relational Database : Example

1 Breaking downdata

2 Typingconstraints

3 Unicityconstraints

4 Integrityconstraints

!"##$%&' !$()*!+,% -%. /0%

!"!#$%& "!' ( )#

!"!)01& "!! ( 2-

!"!)#.& "!! 3 $.

!"##$%&' -1%21)#"# 34526%3)(2# 789

!"!#$%& &*+, - ./.

!"!%-)& ('=E, ) ./.

&4567-

&45670

8'*!"'* 9:&

9:&;<<=,': (=<'&8'*!"'*



Benefits

Relational Database : Benefits

• Data centralisation on the server side

• Constraints allow, in some instances, to avoid data inconsistencies

,! Consistent data

• E�cient : tables containing millions of rows can be easily manipulated

• Querying a correctly structured database allows one to cross-checkdata very rapidly*



Interface

A database requires graphical interface

• Data manipulation in a database is done exclusively with queries,written in SQL (Structured Query Language)

•

!"##$%&' !$()*!+,% -%. /0% -1%21)#"# 34526%3)(2# 789

!"!#$%& "!' ( )# &*+, - ./.

!"!)01& "!! ( 2-

!"!)#.& "!! 3 $.• SELECT t a b l e 1 . t i s s u e ID , t a b l e 1 . TumorType ,

t a b l e 1 . Sex , t a b l e 1 . Age , t a b l e 2 . S t e a t o s i s ,t a b l e 2 . nb adenomas , t a b l e 2 .CRP

FROM t a b l e 1 INNER JOIN t a b l e 2ON ( t a b l e 1 . t i s s u e I D = t a b l e 2 . t i s s u e I D )WHERE t a b l e 1 . T i s sue ID = ’CHC358T ’ ;

• Powerful language, albeit counterintuitive

,! A graphical interface must be associated to the database



Interface

Graphical interface : principle

Mecanism

1 The interface receives instruction from the user, and transform theminto SQL queries sent to the server

2 The server receives the SQL queries, and sends back results

3 The interface receives the results from the server, and displays theresults to the user

Interface types

• 2 types of interface : desktop program or web interface

• In our case, we decided to develop a web interface



Interface

Web interface

• Software installed on a single machine : the web server

• Accessing the interface only requires a web browser

,! Avoids installation and maintenance issues on the client machines

,! Avoids OS compatibility issues (Mac, Windows, Linux, etc. . .)



Interface

Web client / server architecture

��

��

��



1 Introduction



4 CHCDB

5 CHCDBWEB

6 Conclusions



Specificities

Specific features of clinical annotation data

Specificities

• New variables arefrequently added

• Data regarding the samevariable can be inputdi↵erently, depending ofsample provenance andtype (malignant orbenign tumor)

Consequences in a database

• Frequent addition of newcolumns or sub-tables

• Tables contain a lot of columns,with sparsely filled rows

,! Constant maintenance of thedatabase

Clinical annotation data must be stored in a specific database structure :the E.A.V. structure



E.A.V.

Principle

• E.A.V. = Entity Attribute Value[?]

• An E.A.V. is a subset of tables in a relational database, with aspecific organization

• This data organization is particularly suitable of clinical annotationdata

• In the E.A.V., all clinical annotation data is stored in one 3-columnstable :

• Entity : contains the identifier of the entity for which an annotation isstored (In our case, an entity is a tissue)

• Attribute : contains the identifier (e.g. the name) of the annotation variable

• Value : contains the value of the annotation, for a given entity



E.A.V.

Example

!"##$%&' !$()*!+,% -%. /0%

!"!#$%& "!' ( )#

!"!)*+& "!! ( ,-

!"!)#.& "!! / $.

!"##$%&' -1%21)#"# 34526%3)(2# 789

!"!#$%& &012 - .3.



E.A.V.

Example

!"##$%&' !$()*!+,% -%. /0%

!"!#$%& "!' ( )#

!"!)*+& "!! ( ,-

!"!)#.& "!! / $.

!"##$%&' -1%21)#"# 34526%3)(2# 789

!"!#$%& &012 - .3.

256789

256789

':;<=>79

':;<=>79

?@AB>;9

?@AB>;9



E.A.V.

Example

!"##$%&' ?2*"24<%&' ?2<$%

!"!#$%& CBD (

!"!#$%& &>EF;&GHB "!'

!"!#$%& 'IB )#

!"!#$%& C7B@7F9<9 &012

!"!#$%& 5=J@KB5FE@9 -

!"!#$%& !0L .3.

!"!)*+& CBD (

!"!)*+& 'IB ,-

!"!)*+& &>EF;&GHB "!!

!"!)#.& CBD /

!"!)#.& 'IB $.

!"!)#.& &>EF;&GHB "!!

256789 ':;<=>79 ?@AB>;9



E.A.V.

Pros & Cons

Pros

• New annotation = New line inthe table

• No structural modifications

• No sparsely filled tables

Cons

• A lot of lines in the table

• Very complex queries

• Columns are no longer typed

!"##$%&' ?2*"24<%&' ?2<$%

!"!#$%& CBD (

!"!#$%& &>EF;&GHB "!'

!"!#$%& 'IB )#

!"!#$%& C7B@7F9<9 &012

!"!#$%& 5=J@KB5FE@9 -

!"!#$%& !0L .3.

!"!#$%& M;@A!F57;@NBH6F5 &012

!"!)*+& CBD (

!"!)*+& 'IB ,-

!"!)*+& &>EF;&GHB "!!

!"!)#.& CBD /

!"!)#.& 'IB $.

!"!)#.& &>EF;&GHB "!!

!"!)#.& 2KEF59F5 444

?'0!"'0



E.A.V.

Metadata

• Loosinginformationregarding variabledata types isproblematic

,! This data isstored in anancillary table,the metadatatable

!"##$%&' ?2*"24<%&' ?2<$%

!"!#$%& CBD (

!"!#$%& &>EF;&GHB "!'

!"!#$%& 'IB )#

!"!#$%& C7B@7F9<9 &012

!"!#$%& 5=J@KB5FE@9 -

!"!#$%& !0L .3.

!"!#$%& M;@A!F57;@NBH6F5 &012

!"!)*+& CBD (

!"!)*+& 'IB ,-

!"!)*+& &>EF;&GHB "!!

!"!)#.& CBD /

!"!)#.& 'IB $.

!"!)#.& &>EF;&GHB "!!

!"!)#.& 2KEF59F5 444

?2*"24<%&' '212!+,%

CBD ?'0!"'0

&>EF;&GHB ?'0!"'0

'IB 4O&

C7B@7F9<9 PMMQ2'O

5=J@KB5FE@9 4O&

!0L (QM'&

M;@A!F57;@NBH6F5 PMMQ2'O

2KEF59F5 ?'0!"'0



1 Introduction



4 CHCDB

5 CHCDBWEB

6 Conclusions



CHCDB

Software

• R.D.B.M.S. : MySQL• Open source & free• Most widely used open-source R.D.B.M.S.

,! Actively maintained,! Lots of maintenance and development tools

• The machine hosting the R.D.B.M.S. has yet to be bought

Data

• CHCDB’s tables fall into one of three categories• Tissue listings• Clinical annotation data, in the E.A.V. structure• Extraction data


Database structure


1 Introduction



4 CHCDB

5 CHCDBWEB

6 Conclusions



A web interface

Peculiarities

• Installed on the server hosting the R.D.B.M.S.

• Can be reached from any machine on the CEPH network

Features

• Consultation and modificationof clinical annotations for a given tissue

• Listing of tissues and their annotations

• Listing of tissue extractions

• Management (add/modify/delete) of annotation variables

• Batch import of annotations

• Batch import of extraction data


Consultation & modification of the annotations of a given

tissue

Listing of tissues and annotations

Listing of tissue extractions

Annotation variables management


1 Introduction



4 CHCDB

5 CHCDBWEB

6 Conclusions



Missing features in CHCDBWEB

• The tissue management interface is not yet complete

• The batch import interface for annotations and extraction is missing

CHCDB

• Defining a starting set of variables

• Importing existing data into CHCDB

Material

• Acquiring a configuring the machine which will host the database andthe web server

CHCDB and CHCDBWEB should enter production phase in June 2011.


clinical data eav

Technology

tissue extractions listings

concentrations quantities

thomas burguiere

clinical studies

excel les

clinical annotations

malignant tumors

inserm unite 674