Hetionet in Neo4j: the tale of Project Rephetio

Seminar Group on Big and Scientific Data

University of Pennsylvania

February 10, 2017

DSL Conference Room

Moore 102

11:00 am – 12:00 pm

By Daniel Himmelstein

@dhimmel

Slides at slides.com/dhimmel/big-data-seminar

Greene Lab

I'm a data scientist

http://www.greenelab.com/

There are many graph databases. I'm most familiar with Neo4j which is:

  • an ACID-compliant transactional database with native graph storage and processing
  • the most popular graph database according to db-engines.com
  • open source:  Community (GPLv3 licensed) & Enterprise (AGPLv3 licensed) Editions

The Graph Mindset

how Susan Davidson's class was a microcosm of a larger academic network

Graphs are composed of:

  • Nodes
  • Relationships

Nodes / relationships have type:

  • node labels
    (person, course, university)
  • relationship types
    (lecturer, institution)
  • first_name: Daniel
  • last_name: Himmelstein
  • twitter: @dhimmel
  • SSN: 012-34-5678 
  • catalog: CIS550
  • title: Database & Information Systems
  • units: 5
  • catalog: EPID600
  • title: Data Science for Biomedical Informatics
  • units: 1
  • url: www.upenn.edu
  • founded: 1740
  • type: private
  • league: ivy
  • grade: A-
  • grade: B
  • grade: F

What can we do with this graph?

  • Course statistics:
    How many students are in CIS 550?
     
  • Course recommendations:
    What courses do other students in CIS 550 take?
     
  • Course scheduling:
    What room should a course be in so it's nearby other courses that its students are enrolled in?

The Relational Database Model

The Relational Database Model

Limitations:

  1. Relationships require an intermediate table
  2. Schemas are cumbersome to create to maintain

Relationships inherently form graphs

Relational database schema

Graph database schema (metagraph)

Emil Eifrem at GraphConnect 2015

The rise of graph databases

Cypher accelerated graph database adoption

What Cypher looks like

How I became intestested in graphs

http://blog.dhimmel.com/friendship-network/

How do you teach a computer biology?

multilayer network, multiplex network, multivariate network, multinetwork, multirelational network, multirelational data, multilayered network, multidimensional network, multislice network, multiplex of interdependent networks, hypernetwork, overlay network, composite network, multilevel network, multiweighted graph, heterogeneous network, multitype network, interconnected networks, interdependent networks, partially interdependent networks, network of networks, coupled networks, interconnecting networks, interacting networks, heterogenous information network

networks with multiple node or relationship types

A 2012 Study identified 26 different names for this type of network:

hetnet

What's the best software for storing and querying hetnets?

dhimmel/hetio
86
5
2
neo4j/neo4j
42,498
3,071
1,007

GitHub stats from 2016-10-09

  • Hetnet of biology designed for drug repurposing
     
  • ~50 thousand nodes
    11 types (labels)
     
  • ~2.25 million relationships
    24 types
     
  • integrates 29 public resources
    knowledge from millions of studies
     
  • the hardest part:
    licensing of publicly available data

Hetionet v1.0

MetaGraph / Data Model / Schema

Visualizing Hetionet v1.0

Future: all biomedical knowledge in a single network

https://github.com/greenelab/snorkeling

  • Teach computers how to read the literature and extract knowledge.
     
  • Continuously and automatically refine and grow the hetnet.
     
  • Free from any legal restrictions on reuse. 
  • Customized Docker image
  • Digital Ocean droplet
  • SSL from Let's Encrypt
  • readonly mode with a query execution timeout
  • Custom GRASS style
  • Custom guides

Public Hetionet Neo4j Instance

Details at doi.org/brsc

MATCH path =
  // Specify the type of path to match
  (n0:Disease)-[e1:ASSOCIATES_DaG]-(n1:Gene)-[:INTERACTS_GiG]-
  (n2:Gene)-[:PARTICIPATES_GpBP]-(n3:BiologicalProcess)
WHERE
  // Specify the source and target nodes
  n0.name = 'multiple sclerosis' AND
  n3.name = 'retina layer formation'
  // Require GWAS support for the
  // Disease-associates-Gene relationship
  AND 'GWAS Catalog' in e1.sources
  // Require the interacting gene to be
  // upregulated in a relevant tissue
  AND exists(
    (n0)-[:LOCALIZES_DlA]-(:Anatomy)-[:UPREGULATES_AuG]-(n2))
RETURN path

How could multiple sclerosis could affect retina layer formation?

More queries at thinklab.com/d/220

Project Rephetio: drug repurposing predictions

  • Hetionet v1.0 contains:
    • 1,538 connected compounds
    • 136 connected diseases
    • 209,168 compound–disease pairs
    • 755 treatments
  • 1,206 compound–disease metapaths with length ≤ 4
  • machine learning classifier
  • predict the probability of treatment for all 209,168 compound–disease pairs (het.io/repurpose)
  • Project online at thinklab.com/p/rephetio

Systematic integration of biomedical knowledge prioritizes drugs for repurposing
Daniel S Himmelstein, Antoine Lizee, Christine Hessler, Leo Brueggeman, Sabrina L Chen, Dexter Hadley, Ari Green, Pouya Khankhanian, Sergio E Baranzini
bioRxiv. 2016. DOI: 10.1101/087619

Predictions succeed at prioritizing known treatments

Project Rephetio: Does bupropion treat nicotine dependence?

  • Bupropion was first approved for depression in 1985
     
  • In 1997, bupropion was approved for smoking cessation
     
  • Can we predict this repurposing from Hetionet? The prediction was:

Compound–causes–SideEffect–causes–Compound–treats–Disease

Compound–binds–Gene–binds–Compound–treats–Disease

Compound–binds–Gene–associates–Disease

Compound–binds–Gene–participates–Pathway–participates–Disease

MATCH path = (n0:Compound)-[:BINDS_CbG]-(n1)-[:PARTICIPATES_GpPW]-
  (n2)-[:PARTICIPATES_GpPW]-(n3)-[:ASSOCIATES_DaG]-(n4:Disease)
USING JOIN ON n2
WHERE n0.name = 'Bupropion'
  AND n4.name = 'nicotine dependence'
  AND n1 <> n3
WITH
[
  size((n0)-[:BINDS_CbG]-()),
  size(()-[:BINDS_CbG]-(n1)),
  size((n1)-[:PARTICIPATES_GpPW]-()),
  size(()-[:PARTICIPATES_GpPW]-(n2)),
  size((n2)-[:PARTICIPATES_GpPW]-()),
  size(()-[:PARTICIPATES_GpPW]-(n3)),
  size((n3)-[:ASSOCIATES_DaG]-()),
  size(()-[:ASSOCIATES_DaG]-(n4))
] AS degrees, path
RETURN
  path,
  reduce(pdp = 1.0, d in degrees| pdp * d ^ -0.4) AS path_weight
ORDER BY path_weight DESC
LIMIT 10

Cypher query to find the top CbGbPWaD paths

Epilepsy predictions

(browse all predictions at het.io/repurpose)

Discuss at thinklab.com/d/224

Evaluating the top 100 epilepsy predictions

Top 100 epilepsy predictions & their chemical structure

Top 100 epilepsy predictions & their drug targets

Project Rephetio contributions on Thinklab

(see thinklab.com/p/rephetio/leaderboard)

Prior probability of treatment

Methotrexate treats 19 diseases and hypertension is treated by 68 compounds. Methotrexate received a 79.6% prior probability of treating hypertension, whereas a compound and disease that both had only one treatment received a prior of 0.12%.

Questions

https://github.com/cognoma/cognoma

Advertisement: Cognoma Meetup with DataPhilly & Code for Philly

Big Data Seminar at Penn: Hetionet in Neo4j

By Daniel Himmelstein

Big Data Seminar at Penn: Hetionet in Neo4j

Presentation for the Seminar/Reading Group on Big and Scientific Data at Penn (http://www.cis.upenn.edu/~zives/datascience/) on February 10, 2017.

  • 364
Loading comments...

More from Daniel Himmelstein