1. Select traits

  2. Get instruments for traits

  3. Extract SNPs from outcomes

  4. Harmonise data

  5. Perform MR

  6. Upload to neo4j

  7. Add new traits

    • add to existing traits
    • get instruments and find unique instruments
    • extract all outcome SNPs for new trait, and new instruments from other traits
    • harmonise data
    • perform MR
    • upload to neo4j

need two files

  • node information
  • mr estimates
  • heterogeneity stats

neo4j

to run in background:

docker run -d --publish=7474:7474 --volume=$HOME/neo4j/data:/data neo4j:2.3 docker exec -i -t neo4j:2.3 /bin/bash

./neo4j-import
--into mr-eve.db
--id-type string
--nodes:gene ../data/upload/genes.csv
--nodes:snp ../data/upload/snps.csv
--nodes:trait ../data/upload/traits.csv
--relationships:GS ../data/upload/gene_snp.csv
--relationships:GA ../data/upload/snp_trait.csv
--relationships:MR ../data/upload/trait_trait.csv

the paper

the causal map of the human phenome: a first draft

  • intro

    • methods and data enable faster causal inference
    • hypothesis driven analysis should use all methods to scrutinise
    • we can precalculate these estimates but automation requires a number of things still
      • instrument selection using steiger test
      • method selection using linear discriminant analysis
    • here we introduce a comprehensive analysis of 150 traits, introducing two ways for automating instrument selection and method selection
  • results

    • created graph
    • steiger method simulations
    • method selection simulations