TODO.md

In russHyde/coxpresdbr: Parses and Summarises Data from coxpresdb.jp

TODO list for package coxpresdbr

LONG-TERM

• Access coxpresdb via SPARQL

README.md

• explain how to pull out the partners of a single gene

• writes up an example workflow for use of coxpresdbr

• adds a disclaimer re time-to-lookup and suggestion re reducing archive size before running an analysis

R/

coxpresdbr_classes.R

coxpresdbr_workflows.R

• run_coex_partner_workflow( gene_ids, gene_statistics, importer, gene_universe = NULL, n_partners = 100, ...)

  • runs cluster_by_coex_partnership over a CoxpresDbPartners object and appends the resulting graph etc

coxpresdbr_io.R

• get_all_coex_partners(gene_id, importer)

  • uses more than one gene in gene_id argument

coxpresdbr_parse.R

• Warn the user if any of the source genes is absent from the coexpression database

• get_coex_partners

  • rewrite to return a CoxpresDbPartners object

coxpresdbr_stats.R

• .format_coex_edges_for_tidygraph

  • split filtering source/target-ness into a separate function

• cluster_by_coex_partnership

  • add cluster_source_nodes_only = BOOLEAN to formals

  • append to a CoxpresDbPartners object:

    • @partners: data-frame of source_id to target_id where both source_id and target_id are in gene_ids (subset of data returned by get_coex_partners)

• evaluate_coex_partners(x, coex_partners)

  • rewrite to take a CoxpresDbPartners object and return an appended CoxpresDbPartners object.

  • rewrite to take coex_partners as first argument, so that I can pipe from get_coex_partners() %>% evaluate_coex_partners() %>% cluster_by_coex_partnership()

  • set evaluate_coex_partners to a generic function over DGELRT / MArrayLM etc

  • add an alternative-analysis-method switch, eg, evaluate_coex_partners(x, coex_partners, method = c("sumz", "enrichment"))

• evaluate_coex_partners(x: DGELRT/DGEExact, coex_partners)

• evaluate_coex_partners(x: MArrayLM, coex_partners)

• notes

  • What would user want to compare against?

    • Assume the user has performed a separate experiment for which they either have p-values for each gene, or correlation values for each gene against a comparator gene_id

    • Enrichment of gene-partners of a given gene in significant list

    • meta-analysis of p-values / z-scores for the partners of gene_id

    • average correlation of the partners of gene_id with gene_id

    • want some idea of statistical significance, by inverting the averaged z-scores, or by randomly sampling across the gene-universe

  • What data-types would the user have access to?

    • edgeR::DGELRT/DGEExact objects

      • would need to indicate gene_id column of genes component

    • limma::MArrayLM/TestResults objects

      • would need to indicate mappings between rows and gene_id

      • if multiple copies of a gene are present, take the average of it's z-scores before running analysis

    • data-types that can be passed to metap::sumz (p is a vector of one-tailed p-values)

    • data-frame: gene_id, p_value (two-tailed), direction

      • the above datasets should be converted into this form

tests/

test_coxpresdbr_parse.R

• tests both applying filters and combining several source-genes datasets together

test_coxpresdbr_stats.R

• cluster_by_coex_partnership

  • tests for drop_disparities = FALSE

russHyde/coxpresdbr documentation built on Dec. 24, 2019, 11:59 a.m.