notes.md

Notes

Data to use as test/example

• Gene expression time course in human cells (here)
• Sandra's dataset

Features we might want to add

• Options for different distance/dissimilarity measures
• Options for different types of clustering (hierarchical or K-partitioning)
• Produce a "silhouette plot" (http://www.sciencedirect.com/science/article/pii/0377042787901257?via%3Dihub) to diagnose if the chosen number of clusters makes sense
• More interactivity between graphs, like we discussed choosing a gene that is highlighted in the graph, clicking one graph to trigger an action that changes another graph next to it, etc...
• GO enrichment
• motif enrichment
• For Arabidopsis, potential TF binding enrichment using the DAP-seq dataset?
• There is also the following database, which can be useful, for plants TFs: http://plantregmap.cbi.pku.edu.cn/tf_enrichment.php

Already existing apps

What can we add to these already existing apps?

https://kcvi.shinyapps.io/START/

https://www.ncbi.nlm.nih.gov/pmc/articles/PMC5291987/

http://biit.cs.ut.ee/clustvis/

https://asap.epfl.ch/

Interactive heatmaps: https://blog.rstudio.com/2015/06/24/d3heatmap/

Coupled evens in plots: https://plot.ly/r/shiny-coupled-events/

Notes after discussion: focus on clustering and visualising

(Essential functionality is highlighted in bold)

• Input - restrict/allow options based on the input organism

  • if time left: implement solutions for non-model species

• We start with DEG list (generated by some other tool, provide links to tools)

  • supported types of input data: FPKM, TPM, log FC, Z-score
  • data transformation
  • generate heatmap (different methods)
    • Distance metrics: https://bmcbioinformatics.biomedcentral.com/articles/10.1186/1471-2105-15-S2-S2
  • cluster genes by expression patterns
  • support options for clustering (number of clusters, distance measure, etc)
  • silhouete plots to verify/diagnose number of clusters

• Interactivity

  • search by gene name and highlighting it in a row
  • table for genes with name, annotation, expression values
  • support custom annotation for non-model species (gff files)
  • GO enrichment analysis on a cluster - multiple methods/algorithm
  • motif/promoter prediction for clusters (genome sequence required)
  • extract/highlight TFs from the cluster (annotation required) (*)

• Recover results

  • get code to repeat everything with reduced interactivity, buttons for single methods
  • save graphs
  • save table outputs (clustering results + expression + annotation)

• Future work:

  • open an issue and create a branch: issue+N
  • work on one issue at a time

Data