Skip to content

Yale Collab with Aarthi (Smita Krishnaswamy group) where I built signalling knowledge graphs to capture cell communications.

Notifications You must be signed in to change notification settings

FerranC96/sKG4cellCommns

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

76 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

Signalling Knowledge Graphs for Cell Communications

Here I explore an approach for holistic inter- and intra-cellular communication analysis by incorporating literature information as a directed Knowledge Graph, showing that low-dimensional representations of the graph retain biological information and that projected cellular profiles recapitulate their transcriptomes.

This project was undertaken as part of the UCL-Yale exchange programme in collaboration with Smita Krishnaswamy and Aarthi Venkat. Aarthi provided the wavelet module and invaluable help with designing the approach and navigating the issues encountered.

KG design

What we need

  • Capture cell communications from L-R
  • Capture signalling changes via PTMs
  • Summarise above as pathways
  • Who sends, who receives?
  • Compare between cells, clusters, conditions
    • Sc distance metric

Summary

L --> R --> PTMs

Idea of talking (Ligands), hearing (receptors), and listening (PTMs)

Resources

  • NicheNet
  • CellChat
  • REACTOME
  • OmniPath

Methods

  • Pathways should be edge annotations (essentially, annotating the triples with a single/list of pathways)
    • Pathway annotations: From the 6 1st level REACTOME pathways (assuming most/all L, R, PTMS are present in those)
      • 2nd level REACTOME pathway annotations (for visualisation?)
      • 3rd level REACTOME pathway annotations for metadata?
    • Per pathway: Measure GEx/intensity of entities
      • Analyse the 3 distinct layers: Formally define dominance?
        • Ligand dominance -> sender
        • Receiver dominance -> receiver
        • PTM dominance -> ????
      • Summary pathway score:
        • Define cell/group of cells
        • Absolute value: Biased by pathway importance to cell
        • Relative value: (aka EMD with all cells as reference) Relative importance of pathway on different cells
        • Should PTMs provide positive/negative scores? (i.e. inhibitors/activators)
  • KGE: Explore pathway annotations
  • Diffusion method on KG: How the L, R and PTMs relate to one another
    • Wavelets
  • Cells as signals on the graph:
    • Project
    • Compare across cells

Process

  • Handle all 3 sources of data at once: DBs, scRNAseq and cytof
    • Read databases: L and R list
    • Read cytof data: PTMs
  • Build KG: From entities above, define 2 types of relations:
    • L-R cell commns
    • Entity-entity (L to R and R to PTM) interaction if both in the same reactome pathway level
  • Process sc data:
    • Filter data to KG entities and process both modalities
    • All cells from both modalities should add up to same value
  • Downstream analyses
    • Project sc data:
      • Compute distances between cells
        • DR for single modality
        • DR for integrated modalities
      • Cluster cells based on said distance: Functional communication groups
    • Compute pathway scores: Can we compute intensity of signal of cell on each node of the KG? How different would that be from just GEx/AbIntensity
      • Summary pathway score (per cell, per FC cluster)
      • Layer analysis: sender/receiver (per cell, per FC cluster)
    • Inference:
      • In silico pathway ablation -> rebalancing of KG
        • New distance prediction
      • In silico ligand modulation -> simulate ligand presence at diff intensities
        • New distance and pathway score predictions

About

Yale Collab with Aarthi (Smita Krishnaswamy group) where I built signalling knowledge graphs to capture cell communications.

Topics

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published