Knowledge Graphs & Graph Data Science, More Context, Better Predictions - Neo4j at Pharma Data UK 2022

Neo4j, Inc. All rights reserved 2021
1
Knowledge Graphs & Graph Data Science:
More Context, Better Predictions
Dr. Maya Natarajan
Sr. Director, Product Marketing
[Knowledge Graphs]
maya.Natarajan@neo4j.com
Dr. Alexander Jarasch
Life Sciences Technology Specialist
Board Member, HealthECCO
alexander.jarasch@neo4j.com

2
Data Without Relationships Has Very Little Context
2
It’s not the raw numbers, it’s the relationships behind them

The Quest for Human-Like Intelligence & Beyond
3
Graph-Native
ML
Graph
Algorithms
Graph
Queries
Knowledge
Graph Find the patterns
you’re looking for
in connected data
Use unsupervised
machine learning
techniques to
identify associations,
anomalies, and
trends.
Use embeddings to learn
the features in your graph
that you don’t even know
are important yet.
Train in-graph supervised
ML models to predict
links, labels, and missing
data.

4
Building a Knowledge Graph
4
Data Graph
Dynamic Context
Knowledge Graph
Deep Dynamic Context
Semantics
Relationships

5
A New Generation of Complex Data
5
A B C D E
A B C D E
One-to-Many
Relationships
Across Many
Entities
Small, Wide Data Complex Data Hierarchical & Recursive Data
Many-to-Many
Relationships
Nested Tree
Structures
Recursion (Self-
Joins)
Deep
Hierarchies
Link Inference
(If C relates to A and A relates to E,
then C must relate to E)
Node Similarity
Hidden Data
Legacy Data Frozen Data
Legacy SQL Systems Data Lake Fact Tables Graph Data Science - Machine Reasoning
A
C
E

6
Semantics
6
Entity Resolution
& Analysis
Classification
& Tagging
Artificial
Intelligence
C O M P L E X I T Y
Synonym Rings Taxonomy Thesaurus Ontology
Controlled Vocabularies

7
Graph Queries
7
Hey, knowledge graph! Tell me “which genes regulate which pathways”.
MATCH path=(g:Gene)-[r:REGULATES]->(p:Pathway)
RETURN path

8
Graph Data Science Answers the Big Questions
8
What’s important?
What’s unusual?
What’s going to happen next?
But traditional
approaches to data make
it impossible to reveal and
effectively use those
connections as data sizes
become large
Predictive signals get lost in
big data noise
Relationships are
Powerful
Graph Data Science
Uses Relationships to Answer
Critical Questions

9
Graph Algorithms
9
Pathfinding & Search Centrality Community Detection
❏ Shortest Path
❏ Single-Source Shortest Path
❏ Delta Stepping
❏ All Pairs Shortest Path
❏ A* Shortest Path
❏ Yen’s K Shortest Path
❏ Minimum Weight Spanning Tree
❏ K-Spanning Tree (MST)
❏ Random Walk
❏ Breadth & Depth First Search
❏ Degree Centrality
❏ Closeness Centrality
❏ Harmonic Centrality
❏ Betweenness Centrality & Approx.
❏ PageRank
❏ Personalized PageRank
❏ ArticleRank
❏ Eigenvector Centrality
❏ Hyperlink Induced Topic Search (HITS)
❏ Influence Maximization (Greedy, CELF)
❏ Triangle Count
❏ Local Clustering Coefficient
❏ Weakly Connected Components
❏ Strongly Connected Components
❏ Label Propagation
❏ Louvain Modularity
❏ K-1 Coloring
❏ Modularity Optimization
❏ Speaker Listener Label Propagation
❏ K-means Clustering
❏ Leiden Algorithm
❏ Conductance
Heuristic LP Similarity Graph Embeddings
❏ Adamic Adar
❏ Common Neighbors
❏ Preferential Attachment
❏ Resource Allocations
❏ Same Community
❏ Total Neighbors
❏ Node Similarity
❏ Filtered Node Similarity
❏ K-Nearest Neighbors (KNN)
❏ Filtered KNN
❏ Functions: Pearson, Jaccard,
Overlap, Cosine, Euclidean,
Euclidean Distance
❏ Node2Vec
❏ FastRP, FastRPExtended
❏ GraphSAGE

10
Machine Learning Pipelines
10
Node classification:
“What label should this node have?”
Link prediction:
“Is there a relationship between these nodes?”
Labeled data: Pairs of
nodes that are either
linked or not
Features: Pre-existing
attributes, algorithm
results, embeddings
Property Regression (new!)
“What’s the value for this missing property?”
We discover the best model for you - you just supply the data!
Persist and Publish for Production

11
The Drug LifeCycle
11
DISCOVERY DEVELOPMENT COMMERCIALIZATION
⮚ Identify targets to
pursue (traditional or
by drug repurposing)
⮚ Management of
compounds to
identify samples
swaps, HTS gaps,
etc.
⮚ Identify the right
investigator &
patients for the
right trial at the
right time
⮚ Improve patient
outcomes using
patient journey
analytics
⮚ Identify and
connect patient
complaints with
raw material
batches
Target
Discovery
Hit
Generation
Lead
Identification
Lead
Optimization
Animal
Models
Clinical
Trials
FDA/EMA
Review &
Approval
Post
Approval
Manufacturing

12
12
Increasing Volumes of Information
August 2020: 128,000
May 2022: 258, 647

13
Building up the knowledge graph
13

14
Use Case 1
”Freedom-to-Operate on Hemostasis”
Interactive Browsing

15
“Freedom-to-Operate on Hemostasis”
• Based on Gene PF4
known to be
involved in Platelets
Hemostasis
• Browsing the patent
landscape
• Browsing molecular
pathways
• Understanding the
mode of action
15

16
From Pathway to Patents to SysBio-Model in 30s
16

17
Use Case 2
Cross-Species Ontology
Graph Queries

18
Schematic approach
18
Onto
logy
Ter
m
Prot
ein
Tran
scrip
t
Gen
e
Onto
logy
Ter
m
Prot
ein
Tran
scrip
t
Gen
e
HAS HAS
ASSOCIATED
ASSOCIATED
CODES CODES
CODES
CODES
HOMOLOG
SYNONYM
SYNONYM
SYNONYM SYNONYM
SYNONYM
SYNONYM

19
Querying between Ontologies and Genes
19
Example: Equivalent HP-/MP-Term “Deafness”/”Hearing impairment” and ortholog Genes

20
Querying between Ontologies and Genes
20
Example: Equivalent HP-/MP-Term “Deafness”/”Hearing impairment” and ortholog Genes
MATCH (disease:Ontology)-[:HAS_TERM]-
>(hpTerm:Term)<-[:ASSOCIATED_WITH]-
(p1:Protein)<-[:CODES*1..3]-
(t1:Transcript)<-[:CODES*1..3]-
(g1:Gene)-[:HOMOLOG]-(g2:Gene)-
[:CODES*1..3]->(t2:Transcript)-
[:CODES*1..3]->(p2:Protein)-
[:ASSOCIATED_WITH]->(mpTerm:Term)<-
[:HAS_TERM]-(phenotype:Ontology)
WHERE hpTerm ~= ‘deafness’ OR ‘hearing
impairment’
RETURN *

Schematic approach
MATCH (disease:Ontology)-[:HAS_TERM]->(hpTerm:Term)<-[:ASSOCIATED_WITH]-
(p1:Protein)<-[:CODES*1..3]-(t1:Transcript)<-[:CODES*1..3]-(g1:Gene)-
[:HOMOLOG]-(g2:Gene)
-[:CODES*1..3]->(t2:Transcript)-[:CODES*1..3]->(p2:Protein)-
[:ASSOCIATED_WITH]->(mpTerm:Term)<-[:HAS_TERM]-(phenotype:Ontology)
WHERE hpTerm ~= ‘deafness’ OR ‘hearing impairment’
RETURN *
Ont
olo
gy
Ter
m
Pro
tein
Tra
nsc
ript
Ge
ne
Ont
olo
gy
Ter
m
Pro
tein
Tra
nsc
ript
Ge
ne
HAS HAS
ASSOCIATED
ASSOCIATED
CODES CODES
CODES
CODES
HOMOLOG
SYNONYM
SYNONYM
SYNONYM SYNONYM
SYNONYM
SYNONYM

22
Use Case 3
Find The Most Relevant COVID19 Gene
Graph Algorithms

23
• Finding ACE2: the
receptor SARS-Cov2
virus uses to enter the
cell
• GDS PageRank
• Spike protein binds
ACE2
23

24
• Finding ACE2 - the
receptor SARS-Cov2
virus uses to enter the
cell
• GDS PageRank
• Spike protein binds
ACE2
• Target no.1 for vaccines
• Understand the mode of
action
24

25
Use Case 4
Sub-Phenotyping Of Patient Populations
Graph Embeddings

26
Clinical Data from Patients
26
Gender Age BMI diab.status Sample coll #samples #used
surgery disease Histological diagnosis

27
Multi-omics Fingerprint (1) - Lipidomics
27

28
Multi-omics Fingerprint (2) - Transcriptomics
28

29
Transform Data with GDS - Fast Random Projections
29
CALL gds.fastRP.write(
'patients',
{
embeddingDimension: 50,
writeProperty: 'fastrp-
embedding'
}
)
YIELD nodePropertiesWritten

30
k-Nearest Neighbour Clustering (k=5)
30
Graph Algorithms

31
Connect Patient Data with a Knowledge Graph
31

32
Key Messages in this Presentation
In this presentation:
• Knowledge graphs enable the harnessing of relationships and provide
deep, dynamic context important for analytic and machine learning
• Knowledge graphs are the gateway to powerful graph analytics
• Real world pharma use cases:
◦ Use Case 1: Interactive Browsing
◦ Use Case 2: Graph Queries
◦ Use Case 3: Graph Algorithms
◦ Use Case 4: Graph Embeddings
32

33
Thank you!
Contact us:
maya.natarajan@neo4j.com
alexander.jarasch@neo4j.com

Knowledge Graphs & Graph Data Science, More Context, Better Predictions - Neo4j at Pharma Data UK 2022

More Related Content

What's hot (20)

Similar to Knowledge Graphs & Graph Data Science, More Context, Better Predictions - Neo4j at Pharma Data UK 2022 (20)

More from Neo4j (20)

Recently uploaded (20)

Knowledge Graphs & Graph Data Science, More Context, Better Predictions - Neo4j at Pharma Data UK 2022