Concept Whitepaper

This document contains the conceptual background and realization plans for the GePi publication.

Backend

• Use ElasticSearch index with new mapping structure
• Incorporate daily updates
  • Build GePi Pipeline
  • Let run daily

Input

• A Search
• A-B-Search

Types of Input

• Gene IDs
• UniProt IDs?
• Gene Names?

Processing

• Mapping to top homology
• Rather do mapping by gene name?
• Make sure the algorithms do what they are supposed to! Write tests!

Frontend

Pie Chart

Sankey Chart, Simple

• How to deal with enlargening the widget? When the widget is larger, more edges could be shown. But: How many? Depends on number of nodes and the size of the nodes, i.e. the abundances

Sankey Chart, Common Partners

• Same as with simple sankey chart
• How to rank common partner pairs, i.e. which pairs to show?
• Currently: a and b have common partner c -> score(a,b,c) = #(a,c) + #(b,c)
  • downside is that unbalanced hits are not ranked: weighting the score with max(#(a,c) / #(b,c), #(b,c) / #(a,c)) * 2 could be an option (should be in (0,1] )

Table

• Allow sentence filtering by key words
• results per sentence table
  • when hit is in pmc, also deliver pubmed id
• Provide abundance tables
  • A-Search: provide abundance of all interaction partners
  • A-B-Search:
    • show abundance of A and B members, possibly in separate columns
    • show number of different interaction partners for A and B members (potentially including median and/or mean+std)

Usage and Advantage of GePi Capabilities

Use Cases

• Proteomics
• Transcriptomics?
• Pathways - potential interaction partner?

Evaluation

• NatComm scenario against EvexDB
• Evaluate with event-corpora