Pathology is being disrupted by Data Integration, AI & Blockchain

25 June 2018, European Society of Pathology Academy - Belgium
Prof. N Krasnogor - Dr. Yuchun Ding - Dr. Christopher Carey
Interdisciplinary Computing and Complex BioSystems (ICOS) Research Group
http://ico2s.org/
Newcastle Molecular Pathology Node
http://www.newcastlepathnode.org.uk/
School of Computing, Newcastle University
Pathology is being disrupted by Data
Integration, AI & Blockchain
@nkrasnogor

Dr. Jaume Barcadit
Data Analytics, Data Integration,
Computational Intelligence,
Bioinformatics
Prof. Marcus Kaiser
Complex Networks,
Neuroinformatics
Prof. Nat Krasnogor (Director)
Machine Intelligence, Natural & Molecular
Computation, Complex Systems,
Bioinformatics, Systems and Synthetic
Biology, NanoBio-technology
Prof. Anil Wipat (Co-director)
Data Integration, Cloud Computing,
Bioinformatics, Systems and Synthetic
Biology, Gram positive bacteriology
Dr. Phil Lord
Knowledge Management,
Ontologies, Bioinformatics
Dr. Paolo Zuliani
Stochastic Modeling, Formal
Methods, Systems and Synthetic
Biology
Currently 56 individuals:
10 Academics
17 Postdoctoral Researchers
25 PhD Students
3 visiting scientists
1 Administrator
Dr. Angel Goni-Moreno
Synthetic Biology,
Computational modelling
Dr. Harold Fellermann
Complex Systems, NanoBio-technology,
Computational & mathematical modeling
Dr. Yujiang Wang
Computational Neuroscience,
Modeling, complex networks
Dr. Peter Taylor
Computational Neuroscience,
Neuroimage analysis, modeling
Interdisciplinary Computing and
Complex BioSystems (ICOS)
Research Group

Content
• A Perfect Storm or a Sea of Opportunities?
• 1st
Disruptor: The Power of Data Integration
and Systems Thinking
• 2nd
Disruptor: The Power of AI for Image
Understanding
• 3rd
Disruptor: The Power of Blockchain
• Conclusions

Social, Economical and Technological
Trends
5
“Bussolo, Maurizio; Koettl, Johannes; Sinnott, Emily. 2015. Golden Aging : Prospects for Healthy, Active,
and Prosperous Aging in Europe and Central Asia. Washington, DC: World Bank. © World Bank.
https://openknowledge.worldbank.org/handle/10986/22018 License: CC BY 3.0 IGO.”

Trends
6

Trends
7
BMJ 2017; 356 doi: https://doi.org/10.1136/bmj.j41 (Published 13 January 2017)

Trends
• These social and economics trends are strong
incentives to optimise the pathology life cycle
• Advances in data integration & analytics, AI and
blockchain provide strong technological
inducements for optimisation
– Thus: Pathology might be a technological, logistic &
political “low hanging fruit”
– Hence: likely to be disrupted within the next 5 to 10
years.
8

Content
• 1st Disruptor: The Power of Data Integration
• 2nd
Understanding
• 3rd
• Conclusions

The Power of Data Integration and
Systems Thinking
• Typical problem in the life sciences: How to make effective use of
multiple, large-scale, disconnected data sources?
• Typical problem in computer science: How to exploit the strengths of
different algorithms?
 GOAL: Develop new (& extend existing) methods combining diverse
data sources and algorithms

Perturbations in molecular networks disrupt biological pathways
and result in human diseases.
Wang X et al. Briefings in Functional Genomics 2011;10:280-293.
© The Author 2011. Published by Oxford University Press. All rights reserved

Freely accessible at: http://ico2s.org/resources.html
I will describe some of the functionalities and methods in:
Describe network strategies to identify and prioritize gene sets or
functional associations (arising in high throughput experiments)
between a genes/proteins set of interest (target set) and annotated
genes/proteins sets (reference set)
Describe network strategies to identify and prioritize gene sets or
functional associations (arising in high throughput experiments)
between a genes/proteins set of interest (target set) and annotated
genes/proteins sets (reference set)
In collaboration with:
Dr. Enrico Glaab (UoN and currently at
University of Luxembourg )
Dr. Anais Baudot (Unversite d’Aix-
Marseille)
Prof. Reinhard Schneider (University of
Luxembourg )
Prof. Alfonso Valencia (CNIO )
Prof. Doron Lancet (Weizmann Institute)
In collaboration with:
Dr. Enrico Glaab (UoN and currently at
University of Luxembourg )
Dr. Anais Baudot (Unversite d’Aix-
Marseille)
Prof. Reinhard Schneider (University of
Luxembourg )
Prof. Alfonso Valencia (CNIO )
Prof. Doron Lancet (Weizmann Institute)

What is TopoGSA?
TopoGSA is a web-application mapping
gene sets onto a comprehensive human
protein interaction network and analysing
their network topological properties.
Two types of analysis:
1. Compare genes within a gene set:
e.g. up- vs. down-regulated genes
2. Compare a gene set against a
database of known gene sets
(e.g. KEGG, BioCarta, GO)
TopoGSA: Network topological
analysis of gene sets

TopoGSA - Methods
• the degree of each node in the gene set
• the local clustering coefficient Ci for each node vi in the gene set:
where ki is the degree of vi and ejk is the edge between vj and vk
• the shortest path length between pairs of nodes vi and vj in the gene set
• the node betweenness B(v) for each node v in the gene set:
here σst(v) is the number of shortest paths from s to t passing through v
• the eigenvector centrality for each node in the gene set
TopoGSA computes topological properties for an uploaded gene set
and matched-size random gene sets
TopoGSA computes topological properties for an uploaded gene set
and matched-size random gene sets

LEGEND:
• Cellular processes
• Environmental information processing
• Genetic information processing
• Human diseases
• Metabolism
• Cancer genes
General results:
• Metabolic pathways
have high shortest path
lenghts and low bet-
weenness
• Disease pathways and
cancer gene sets tend to
have high betweenness
and small shortest path
lenghts
Mean node
betweenness
Mean clustering
coefficient Mean shortest
path length
PPI network  MIPS, DIP, BIND, HPRD and InAct
9393 proteins and 38857 interactions
Also for yeast, fly, worm and arabidopsis
PPI network  MIPS, DIP, BIND, HPRD and InAct
9393 proteins and 38857 interactions
Also for yeast, fly, worm and arabidopsis

Main dataset
QMC breast cancer microarray data set
• Platform: Illumina Sentrix Human-6 BeadChips
• Pre-normalized data (log-scale, min: 4.9, max:
13.3)
• 128 samples and 47,293 genes
• 3 tumour grades: 1 (33), 2 (52), 3 (43)
• Probe level data analysis: Bioconductor
beadarray package
• Public access to data set:
http://www.ebi.ac.uk/microarray-as/ae
accession number: E-TABM-576
grade1 grade 3
Heat map: 30 most differentially
expressed genes vs. samples (grade
1 and grade 3)
genes

Expression level and topology
Heat map: 50 most significant genes Box plot: 4 most significant genes
QMC Breast cancer data
Expression levels across 3 tumour grades:
STK6 MYBL2
KIF2C AURKb
www.arraymining.netwww.topogsa.net

Expression level and topology
Topological analysis of Selected Genes
• Results of within-gene-set comparison:
Estrogen receptor 1 gene and apoptosis regulator Bcl2, both up-regulated
in luminal samples, have outstanding network topological properties (higher
betweenness, higher degree, higher centrality) in comparison to other genes.
• Results of comparison against reference databases:
- Metabolic KEGG pathways are most similar to the uploaded gene set in
terms of network topological properties.
- Most similar BioCarta pathways: Cytokine, Differentiation and inflammatory
pathways.

Idea:
Enlarge pathways by adding
genes that are “strongly
connected‘‘ to the
pathway-nodes or
increase the pathway‘s
“compactness“
PathExpand: Expanding pathways and
cellular processes

• Our procedure extended 159 pathways from
BioCarta, 90 from KEGG and 52 from
Reactome.
• The pathway sizes increased on average from
113% to 126% of the original size.
• Our procedure extended 159 pathways from
BioCarta, 90 from KEGG and 52 from
Reactome.
• The pathway sizes increased on average from
113% to 126% of the original size.
Validation shows:
1)The proteins added are well connected and central in the protein interaction
network
2)The added proteins display gene ontology annotations matching better to the
original cellular pathway/process annotations than random proteins
3)Are enriched in processes known to be related to cellular signaling
4)Our method is able to recover known cellular pathway/process proteins in a
cross-validation experiment
Example case: BioCarta BTG family proteins and cell cycle regulation
Black: Original pathway nodes – Green: Nodes added based on connectivity
Added cancer gene

Pathway enlargement – Example 1
Alzheimer disease pathway
•More than 20 proteins annotated in our
PPIN
•5 added proteins by the extension
process
•3 known disease associated
•2 candidates: METTL2B, TMED10

Pathway enlargement – Example 2
•Complex signaling system,
sharing intracellular cascades
•New regulators
•New crosstalk proteins
Interleukin signaling pathways

Target
Set
Target
Set
Reference Set
1
Reference Set
1
Reference Set
n
Reference Set
n
2) Feed
1) Produce 3) Filter/Compare/
Overlap/Etc
4) Transfer
5) New Hypothesis/Experiments/Insights
Multiple Approaches for “Enrichment Analysis”:
•Over-representation analysis (ORA)
•Gene set enrichment analysis (GSEA)
•Integrative and modular enrichment analysis (MEA)
Huang, D. et al., Bioinformatics enrichment tools: paths toward the
comprehensive functional analysis of large gene lists. Nucleic Acids Res.,
37 (1), 2009
Network-based Functional Association Ranking

Key limitations of previous “Enrichment
Analysis”:
1.ORA techniques often have low discriminative power with scores varying
widely with small changes in the overlap size.
2.Functional information captured in the graph structure of a molecular
interaction network connecting the gene/protein sets of interest is disregarded.
3.Genes and proteins in the network neighborhood, in particular those with
missing annotations, are not taken into account.
4.The recognition of tissue-specific gene/protein set associations is often
statistically infeasible.

Key idea behind EnrichNet
Target SetTarget Set
Gene1
.
.
.
GeneN
Gene1
.
.
.
GeneN
Reference Set 1Reference Set 1
Reference Set nReference Set n
Gene1
.
.
.
GeneN
Gene1
.
.
.
GeneN
Protein1
.
.
.
ProteinM
Protein1
.
.
.
ProteinM
∪∩Target SetTarget Set
Gene1
.
.
.
GeneN
Gene1
.
.
.
GeneN
Reference Set 1Reference Set 1
Reference Set nReference Set n
Gene1
.
.
.
GeneN
Gene1
.
.
.
GeneN
Protein1
.
.
.
ProteinM
Protein1
.
.
.
ProteinM
∪

Input:
– 10 or more human gene or protein identifiers
– Selection of reference database, e.g., KEGG, BioCarta, WikiPathways, Reactome, PID, Interpro,
GO.
Processing:
– Maps target and reference sets to genome scale molecular interaction networks
• Two default ones available
• User can provide her/his own
– RWR to calculate distance between target and reference sets mapped into the large network
– Comparison of these scores against a background model
Output:
– Ranking table of reference dataset
• cellular pathways, processes and complexes
– Network and Tissue (60) specific scores
– For each pathway  interactive visualization of its embedding network
– Zoom, search, highlight, retrieve annotation/topological data
What does it do?

How does it work?
Distances (~F(Pt
)) from target nodes
to reference ones is calculated by a
Random Walk with Restart (RWR)
procedure.
These distances are then linked to a
background model…
Distances (~F(Pt
)) from target nodes
to reference ones is calculated by a
Random Walk with Restart (RWR)
procedure.
These distances are then linked to a
background model…
Connected human interactome graph
derived from STRING 9.0 database
Edges weighted by the STRING
combined confidence score
normalized to range [0,1])
Connected human interactome graph
derived from STRING 9.0 database
Edges weighted by the STRING
combined confidence score
normalized to range [0,1])

Results: Identification of novel functional associations
The network-based Xd-score ranking identifies several new associations missed by the
classical approach (ORA)
We use dataset examples pairs such that:
•target gene sets are all mutated in different diseases without additionally available
expression level data  they cannot be analyzed with “expression-aware” gene set
enrichment analysis techniques
•zero or insignificant overlap size  receive low ORA scores (Q-values> 0.05) but Xd-
scores above the significance threshold obtained from the linear regression fit
•these results point to functional associations that reflect dense networks of
interactions between the target and reference datasets  overlooked by approaches

• Largest connected component for the network structure obtained when
comparing the gastric cancer mutated gene set against the pathway Role of
Erk5 (Extracellular signal-related kinase 5) in Neuronal Survival (h_erk5Pathway)
from the BioCarta database, describing a signalling cascade which induces
transcriptional events promoting neuronal survival.
• These datasets have an intersection of only three genes (HRAS, NRAS & KRAS)
and would therefore not have been considered as significantly associated ORA.
• The network-based Xd-score (0.26, which is a significant threshold VS only 0.08
Q-value in ORA with Fisher test), highlights functional associations.
• These reflect abundance of molecular interactions between the corresponding
proteins for these gene sets and their shared network neighborhoods.
• This dense network of interactions corroborates previous findings linking
extracellular signal-related kinases (ERKs) to gastric cancer via an induction of
the putative tumor suppressor gene DDMBT1 (deleted in malignant brain
tumors 1) by a reduced ERKs activity.

• Largest connected component for the network structure obtained
when comparing two datasets, bladder cancer mutated genes and
the genes for the Gene Ontology (GO) term ‘tyrosine
phosphorylation of Stat3
• These share only a single gene (NF2)  no association can be
inferred from ORA
• The high Xd-score for this gene set pair (0.80) points to a functional
association via multiple connecting molecular interactions, which is
confirmed by the visualization.
• In agreement with the previous observation that the down-
regulation of STAT3 phosphorylation by means of silencing the Rho
GTPase CDC42 is linked to the suppression of tumour growth in
bladder cancer.
• Rho GTPases like CDC42 are known to frequently participate in
carcinogenic processes
• Their involvement in bladder cancer is also reflected by a high Xd-
score of 0.71 for the GO biological process ‘regulation of Rho GTPase
activity’ (GO:0032319), which also shares only one gene with the

c) • Largest connected component for the network structure
obtained when comparing two datasets, genes implicated in
Parkinson’s disease (PD) and the ‘regulation of interleukin-6
biosynthetic process’ from the Gene Ontology database
• A significant XD-score (0.77, significance threshold: 0.73) even
when sharing only one gene (IL1B)
• The corresponding sub-network reveals a dense cluster of
interactions that interlink the PD gene set with the interleukin-
6 pathway.
• This corroborates previously identified links between PD and
inflammation and reports of elevated levels of interleukin-6 in
the cerebrospinal fluid of PD patients

Content
• 2nd Disruptor: The Power of AI for Image
Understanding
• 3rd
• Conclusions

Where Do We Stand?
34
From “Brief History of Artificial Intelligence” by R. Buest (http://analystpov.com/strategy/figure-a-brief-history-of-artificial-intelligence-25689)

Lots of Rapid Progress
35

36
“How Is Artificial Intelligence Disrupting Healthcare?” by M. Matthews (http://hin.com/blog/2017/06/19/infographic-how-is-artificial-intelligence-disrupting-healthcare/)

Raw ImageRaw Image Processed ImageProcessed Image Initial SimulationInitial Simulation

Where Could AI Help?
38

Challenge 1:
• NHS pathologists spend hours everyday to
analyse IHC slides.
• Switching between multiple IHC stained
slides to look at the corresponding
locations is time consuming and lab
intensive.

Solution: Automated Colocation Analysis
• We developed an automated solution for
the alignment of differently stained
sections, to enable concurrent analysis of
multiple IHC markers.
• 2.5 minutes for each case with 4 antibodies
• Accurate when the tissue samples are in
good quality.

Automated Image Alignment To Analyse Whole Slide Image of Lymph Node
CD30 (aligned) CD68 (aligned)PAX5 (aligned)
H&E (original)

Challenge 2:
• Most slides for histopathological
examination were H&E only stained slides.
• Automated analysis on histology sample
with complex structures is challenging due
to their large variances in appearance
(shape, size, and texture)

Solution: Digital Pathology & AI
• We developed a user trainable software solution,
DeepButton, for analysing digital microscopy images.
• DeepButton uses pre-trained deep learning models to
provide easy-to-use solutions for automating the
segmentation of difficult images in digital pathology.

DeepButton
provides an out-of-
the-box software
with an easy-to-use
user interface to
perform batch
image segmentation
with a single click of
a button

Typical raw images and training data

Raw image and result

AI can be trained quickly
mouse hippocampus colonic crypts cancer (breast) nuclear

AI can be trained for one task
but works on others
H&E nuclei IHC nuclei
Trained with
diffuse large B-cell lymphoma
Picks up IHC diffuse mouse brain
cell nuclei

AI can be trained quickly for very
challenging task
Breast cancer cell nuclei
(H&E estrogen receptor
positive breast cancer)

Sometimes AI does better job
human
Quiz: a human expert and the AI have both annotated colonic
crypts in pink, but which is which??

human
Human expert AI

than commercial software
CommercialCommercialAIAI

AI can be trained easily on different cases
Cancer metastasisEpithelial structureEpithelial structure Surgical tool Polyps Angiodysplasia

Metastases in H&E stained lymph
node

Challenge 3: Post-image
analysis
• Statistical analysis are usually performed
with computer scripts.
• User interface for the existing software are
very complicated and difficult to use.
• There is very limited options to analyse
multiplexed images.

Typical Multispectral Image Structure
(Vectra 3.0)

Solution: DeepButton Analyst
• DeepButton Analyst is a software aims to
deliver a out-of-box solution to perform
image visualisation, segmentation,
quantitative analysis, proximity graph, on
brightfield images, multiplexed images,
fluorescence images in a simplified user
interface.

Multiplexed image of a whole tissue

Segmentation overlay of a whole
tissue

Cell network distribution
Topological analysis reveals a PD-L1-associated
microenvironmental niche for Reed-Sternberg
cells in Hodgkin lymphoma.
http://www.bloodjournal.org/content/130/22/2420

Challenge 4: Histology in 3D
• Histology analysis are usually performed on
2D serial sections. A lot of information is
missing.
• Not much available to analyse stacks of
histology serial sections.

Solution: DeepButton Analyst
• DeepButton Analyst is integrated with
multiple next generation features to
perform 3D alignment, reconstruction and
visualisation with a stacks of digital slides.

Using the tool (demo)
https://youtu.be/MF5xhvv1Xpc?t=16s

Content
• 2nd
Understanding
• 3rd Disruptor: The Power of Blockchain
• Conclusions

Why Blockchain in Pathology?
• Biosamples and their data-footprint
– Who collects?
– Who handles?
– Who shares?
– Who owns?
– Who benefits?
• Data lakes and outdated privacy protocols
impede progress that would benefit patients,
doctors and national health services

Blockchain explained (1)
blockchain is a ledger
but
A digital,
cryptographically
signed & decentralised
one.
Picture from PBS, copyright unkown

blockchain stores only digital
elements
•Sign with private key: clear
ownership
•Verify with public key
•Features:
• Immutable
• No central authority
• Secure and transparent
• Private
• Scalable and Available
•Coins as unit/currency for
operation From Nakamoto 2008

blockchain ledger is
distributed
•Each miner keeps a
copy of the ledger
•Peer-to-peer protocol
to distribute updates
•Locally “trust-less”,
collectively
“trustworthy”
From Coindesk.com

Because blockchain is distributed, miners need to
reach consensus about all updates and verify them:
•Group transactions in “blocks”
•Use a consensus algorithm to agree veracity of
transaction
•Proof of Work
– Need to invest (CPU cycles), for a reward 
miner gain a stake in the blockchain
– Introduces a competition, that no single
party can always win
From inerciatech.com
From www.processexcellencenetwork.com

Ethereum = Blockchain + Smart Contracts
• General blockchain platform for a large set of
distributed applications (daps)
– Turing complete (little) programs
– Smart contracts can call smart contracts
– Executed when specified conditions are satisfied in
blockchain, e.g., monthly payment, items
delivered/received, etc
– Execution output made available in blockchain, e.g.
mortgage agreement
b b+1 b+2 b+3 b+4 b+5
Cond1 Cond2 Cond3 Output
SC
1
2
3
Cond1
Cond2
Cond1
Output

Why Blockchain in Pathology?
• Biosamples and their data-
footprint
– Who collects?
– Who handles?
– Who shares?
– Who owns?
– Who benefits?
• Data lakes and outdated privacy
protocols impede progress that
would benefit patients, doctors
and national health services
Could be processed and controlled via Ethereum
(or Neo) smart contracts on Blockchain
Blockchain offers the opportunity
to build the economic incentives
to overcome current obstacles

• Example applications of (simple) network-based scoring methodology
• Illustrate the utility of the approach for identifying novel functional associations
between gene/protein sets and disease states
• Reflect known direct and indirect molecular interactions between their members
rather than only the size of their overlap
• Ever larger datasets are being made available but they are usually difficult to integrated and inter-
operationalize
• Difficulty in interpreting in the clinic
Conclusions: 1st Disruptor - The Power of Data Integration and
Systems Thinking
Freely accessible at: http://ico2s.org/resources.html
This work was supported by the UK’s Biotechnology and Biological Sciences Research Council [BB/F01855X/1], the Engineering and Physical Sciences Research Council (EP/J004111/1), the Spanish Ministry for Education and Science [BIO2007-66855] and a Morris Belkin visiting professorship.

Conclusions: 2nd Disruptor - The Power of AI for Image
Understanding
• AI is (sort of) coming of age
• Deep Button is a concept:
• Talk to us so we can build
something that is useful to you

(Inter)National cross-hospitals collaborations to procure, transact,
license and shepherd large quantity of patient derived data.
RVIRVI
Free
man
Free
man
QMCQMC
GOSGOS
WPHWPH
GOSGOS
Own patients
Own images
Own annotation
Data
Own patients
Own images
Own annotation
Data
Own patients
Own images
Own annotation
Data
Own patients
Own images
Own annotation
Data
Own patients
Own images
Own annotationData
Own patients
Own images
Own annotation
Data
Push data to the blockchain
Pull Knowledge & tokens
from the blockchain
Conclusions: 3rd
Disruptor – The Power of Blockchain
The
(inter)natio
nal health
service
blockchain
The
(inter)natio
nal health
service
blockchain

Acknowledgements
•Prof. Aad van morsel for the blockchain slides
•Prof. P. Sloan, Prof. N. Reinolds, Dr. C. Bacon, Dr. C. Lamb of NMPN
•EPSRC, BBSRC and MRC for funding
•The organisers of ESPA

Pathology is being disrupted by Data Integration, AI & Blockchain

Recommended

Recommended

More Related Content

What's hot

What's hot (20)

Similar to Pathology is being disrupted by Data Integration, AI & Blockchain

Similar to Pathology is being disrupted by Data Integration, AI & Blockchain (20)

More from Natalio Krasnogor

More from Natalio Krasnogor (20)

Recently uploaded

Recently uploaded (20)

Pathology is being disrupted by Data Integration, AI & Blockchain

Editor's Notes