Integrate all: hetnets in human disease

November 20, 2015

Asilomar Grounds

dhimmel on:

QBC Retreat

—Daniel Himmelstein

Baranzini Lab @ UCSF

Sandler Neurosciences Center

Sergio

PhD in Biological & Medical Informatics

Founding Insight

  • context
    bioinformatics → data explosion
     
  • goal
    mine the data to advance human health
     
  • problem
    high-throughput data tends to predict     weakly
     
  • remedy
    combine diverse datasets into a strong predictor
     
  • method
    heterogeneous network (hetnet) edge prediction

Predicting disease-associated genes

network of pathogenesis:

 

  • integrate diverse data to provide context
     
  • 18 metanodes
    40,343 nodes
     
  • 19 metaedges
    1,608,168 edges

type is essential when operating on hetnets

metapath-based approach

feature extraction: the DWPC

learning to classify:

regularized logistic regression

mechanisms of pathogenesis:

comparing gene set collections

extras

mechanisms of pathogenesis:

comparing metapaths

Predicting withheld MS associations

Novel MS associations

performance and permuation

Metagraph for predicting drug repurposing

Network for drug repurposing

  • 50k nodes
    10 types
     
  • 3M edges
    27 types
     
  • 28 public resources
     
  • open for reuse & reproducibility

1. copyright

limited by:

  • fair use

  • originality (excludes facts)

2. contract

agreement entered into to receive access to a resource

  • can impose restrictions beyond copyright

restrictions on data

automatically granted to "original works of authorship" giving the exclusive right to:

  • copy​​

  • distribute

  • create derivatives

1. ∅ license

3. ∅ distribute

  • MSigDB — publicly-funded project from the Broad
  • publication data supplements

complications

4. standard

  • 9 resources
  • all rights reserved
  • upon contact:
    • 1 permission
    • 0 licenses added

2. unclear

  • 4 resources
  • clarification after laborious and slow permission requests
  • 11 resources
  • incompatibilities

5. government

  • 4 resources
  • public domain

Resolution after months & 5000+ word discussion: mixed approach

Recommendation:

release data as CC0

(public domain)

massively collaborative open science

22 reviewers, 53 discussions, 293 comments

stats from 2015-10-20

Mining knowledge from 69 years of biomedical publication

  • MEDLINE: curators annotate paper topics
     
  • 21 million articles
  • since 1946
  • 5,594 journals
     
  • cooccurence of two topics indicates a relation
  • disease–symptom cooccurence in 363,928 articles, 696,252 for disease–anatomy

Discuss on

anatomy

symptom

Mining MEDLINE for disease context

LINCS L1000: ~20,000 small molecules

We combine all signatures for each DrugBank compound to get a consensus signature

Brueggeman

transcriptional signatures discriminate

diuretics (DR) & anti-Inflammatories (AA)

← genes →

Disease-specific models uncover therapeutic signatures

  • MEDI-HPS:
    • RxNorm
    • MedlinePlus
    • SIDER 2
    • Wikipedia
  • ehrlink: linked data from health records
  • LabeledIn: expert and MTurk curated drug labels
  • PREDICT:
    • UMLS links
    • drugs.com
    • drug labels

catalog of indications

aggregated 4 databases:

yielding 1,388 indications

drug disease ajg csh eq
Acetylsalicylic acid systemic lupus erythematosus DM DM 1
Alprazolam systemic scleroderma SYM SYM 1
Baclofen multiple sclerosis SYM SYM 1
Bupropion panic disorder SYM DM 0
Captopril rheumatoid arthritis NOT NOT 1
Cisplatin hematologic cancer DM DM 1
Cladribine hematologic cancer DM DM 1
Clopidogrel coronary artery disease DM DM 1
Cocaine dental caries NOT SYM 0
class ajh csh
DM 26 32
SYM 20 17
NOT 4 1

curation pilot results

  • ~58% disease modifying
  • ~37% symptomatic
  • ~5% non-indications
  • discuss on 

66% ✓

metapath nonzero auroc
CcSEcCiD 0.590 0.897
CiDiCiD 0.255 0.840
CiDaGaD 0.411 0.820
CtGtCiD 0.227 0.807
CiDpSpD 0.410 0.801
CiDlAlD 0.400 0.797
CtGiGaD 0.335 0.710
CuG<kuGaD 0.425 0.692
CtGvD 0.009 0.514

Results for all metapaths ≤ length 3

feature performance

(including symptomatic indications)

common side effects

shared indications

shared gene associations

shared targets

shared symptoms

knowledge-bias free

genetics

  • graph database
     

  • designed for property graphs
    natively supports hetnets
     

  • cypher query language

MATCH path = (source:Disease)--(:Symptom)--(target:Disease)
WHERE 
source.name = 'multiple sclerosis' AND
target.name = 'psoriasis'
RETURN path

Common symptoms between MS and psoriasis

MATCH path = 
(source:Compound)-[:TARGET|BINDING]-
(:Gene)-[:VARIATION]-(target:Disease)
WHERE target.name = 'multiple sclerosis'
RETURN path

Compounds that target MS-associated genes

DWPC in Cypher

Made with Slides.com