This document discusses fostering serendipity through linking large biomedical datasets. It linked over 30 billion triples from The Cancer Genome Atlas (TCGA) and over 23 million publications from PubMed. It developed an architecture called TopFed to continuously integrate new data through parallel querying. TopFed was evaluated against the FedX system and shown to have significantly better performance, with query runtimes over 75 times faster for some queries. A visualization interface was also created to explore the linked data.

1. Fostering Serendipity through Big
Linked Data
Muhammad Saleem , Maulik R. Kamdar , Aftab Iqbal ,
Shanmukha Sampath , Helena F. Deus , and Axel-Cyrille
Ngonga Ngomo
Semantic Web Challenge at ISWC2013, October 21-25 , 2013, Sydney, Australia

2. Agenda
• Motivation
• Datasets
• Architecture
• Evaluation
• Requirements
• Demo
• Conclusion and Future Work

3. Motivation
Fostering Serendipity through Big Data
Triplification, Continuous Integration,
and Visualization

4. Triplification: Linked TCGA
• TCGA is publicly accessible atlas of cancer
related data from National Cancer Institute
(NCI)
– 9000 patients
– 33 cancer types
– 147,645 raw data files
– 12.7 TB
• Only 46% of the total expected data with
new data being submitted every day
• Goal is to enable cancer researchers to
make and validate important discoveries
• Total Linked TCGA > 30 billion triples
(Largest Dataset of LOD)

5. Triplification:PubMed
• Collection of publications from the bio-medical
domain
• Large amount of metadata (MESH Terms)
• 23+ million publications
• 10,000 new publications/month

6. Big Data Continuous Integration
TopFed
Parser
Federator Optimizer
Integrator
Results
SPARQL Query Results
Sub-query
PubMed
Entrez Utilities
RDFizer
Auto
Loader
TCGA Data
Portal
SPARQL
endpoint
RDF
SPARQL
endpoint
RDF
SPARQL
endpoint
RDF
Index

7. Exon-Expression
Methylation
C-1 ∨ Category
Colour = blue
For each query triple t(s, p, o) ∈ T
Highly Scalable
b1 b2 p1 p2 p3 p4 p5 p6 g1 g2 g3 g4 g5 g6 g7 g8 g9
C = {CNV, SNP, E-Gene, E-Protein, miRNA, Clinical}
M = {beta_value, position} F = {Expression-Exon}
(CNV, SNP, E-Gene,
miRNA,
E-Protein, Clinical)
D = {seg_mean, rpmmm, scaled_est, p_exp_val}
B = {DNA-Methylation}
C-1 = {{p ∈ {D ∪ A ∪ G} ∨ {p = rdf:type ∧ o ∈ C}} ∧ {{S-Join(p, D ∪ C) ∨ P-Join(p, D ∪ C) } ∨ {!S-Join(p, M ∪ B ∪ E ∪ F) ∧
!P-Join(p, M ∪ B ∪ E ∪ F) }}}
C-2 = {{p ∈ {E ∪ A ∪ G} ∨ {p = rdf:type ∧ o ∈ F}} ∧ {{S-Join(p, E ∪ F) ∨ P-Join(p, E ∪ F)} ∨ {!S-Join(p, M ∪ B ∪ D ∪ C) ∧
!P-Join(p, M ∪ B ∪ D ∪ C) }}}
C-3 = {{p ∈ {M∪ A} ∨ {p = rdf:type ∧ o ∈ B}} ∧ {{S-Join(p,M ∪ B) ∨ P-Join(p, M∪ B) } ∨ {!S-Join(p, E ∪ F ∪ D ∪ C) ∧
!P-Join(p, E ∪ F ∪ D ∪ C) }}}
IF tumour lookup is successful
forward to corresponding
leaf
Else
broadcast to every one
A = {chromosome, result, bcr_patient_barcode} G = {start, stop}
E = {RPKM}
Tumours
SPARQL
endpoints
C-2 ∨ Category
Colour = pink
C-3 ∨ Category
Colour = green
1-16 17-33 1-5 6-11 12-16 17-22 23-27 28-33 1-4 5-8 9-12 13-16 17-20 21-24 25-27 28-30 31-33

8. Evaluation:Number of Sub-Query Submission
60
50
40
30
20
10
FedX number of Sub-Query Submission TopFedE number of Sub-Query Submission
• TopFed number of sub-queries submission is 1/3 to FedX
• Number of ASK requests
– FedX 480
– TopFed 10
0
1 2 3 4 5 6 7 8 9 10 Avg

9. Evaluation: Query Runtime
100000
10000
1000
100
10
1
1 2 3 4 5 6 7 8 9 10 Average
Query Execution Time (msec) in
log scale
FedX TopFed
• TopFed outperform FedX significantly on 90% of the queries
• On average, the query run time of TopFed is about 1/3 to that
of FedX
• TopFed‘s best run-time (query 2, query 3) is more than 75 times
smaller than that of FedX

10. Big Data Track Requirements
• Data Volume
– 7.36 billion triples from Linked TCGA
– 23 million publications from PubMed
• Data Variety
– The Linked TCGA data was extracted from raw text files of different
structures
– Processed the metadata associated with PubMed publications and
transform them into RDF
– Unstructured data (publication abstracts) is processed to extract
mentions of gene names and cancers
• Data Velocity
– TCGA data doubles /2 months
– PubMed publications 10k/month

11. Big Data Visualization

12. Tumor-wise Visualization

13. PubMed Paper-wise Visualization

14. Genome-wise Patients Results Visualization

15. Everything is Public
• Demo: http://srvgal78.deri.ie/tcga-pubmed/
• TopFed: https://code.google.com/p/topfed/
• TCGA Data Refiner, RDFizer: http://goo.gl/vSnBEJ
• Utilities: http://goo.gl/kNrFdI
• Linked TCGA : http://tcga.deri.ie/
saleem@informatik.uni-leipzig.de
AKSW, University of Leipzig, Germany