In vjcitn/msigTextM: text mining for msigdb sets construed as documents

library(ssrch)
library(msigTextM)
library(DT)

Introduction

MSigDb is a curated collection of gene sets, with enumerations of gene symbols or ENTREZ identifiers in sets that occupy higher level categories.

The main categories (and subcategories, indented) are

• H: hallmark
• C1: positional
• C2: curated
  • CGP: chemical and genetic perturbations
  • CP: canonical pathways (also subclassed by source)
• C3: motif
  • MIR: microRNA target sets
  • TFT: transcription factor target sets
• C4: computational
  • CGN: cancer gene neighborhoods
  • CM: cancer modules
• C5: Gene Ontology
  • BP: (biological process)
  • MF: (molecular function)
  • CC: (cellular component)
• C6: oncogenic
• C7: immunologic

Here we tabulate the counts of gene sets by category and sub-category.

             main category
subcategory   ARCHIVED   C1   C2   C3   C4   C5   C6   C7    H
                     0  326    0    0    0    0  189 4872   50
  BP                 0    0    0    0    0 4436    0    0    0
  C5_BP            527    0    0    0    0    0    0    0    0
  C5_CC            159    0    0    0    0    0    0    0    0
  C5_MF            178    0    0    0    0    0    0    0    0
  CC                 0    0    0    0    0  580    0    0    0
  CGN                0    0    0    0  427    0    0    0    0
  CGP                0    0 3433    0    0    0    0    0    0
  CM                 0    0    0    0  431    0    0    0    0
  CP                 0    0  252    0    0    0    0    0    0
  CP:BIOCARTA        0    0  217    0    0    0    0    0    0
  CP:KEGG            0    0  186    0    0    0    0    0    0
  CP:REACTOME        0    0  674    0    0    0    0    0    0
  MF                 0    0    0    0    0  901    0    0    0
  MIR                0    0    0  221    0    0    0    0    0
  TFT                0    0    0  615    0    0    0    0    0

The vocabularies of the gene set collections are important for many aspects of utilization.
STANDARD_NAME attributes for C7 gene sets look like

 [1] "KAECH_NAIVE_VS_DAY8_EFF_CD8_TCELL_UP"          
 [2] "KAECH_NAIVE_VS_DAY8_EFF_CD8_TCELL_DN"          
 [3] "KAECH_NAIVE_VS_DAY15_EFF_CD8_TCELL_UP"         
 [4] "KAECH_NAIVE_VS_DAY15_EFF_CD8_TCELL_DN"         
 [5] "KAECH_NAIVE_VS_MEMORY_CD8_TCELL_UP"            
 [6] "KAECH_NAIVE_VS_MEMORY_CD8_TCELL_DN"            
 [7] "KAECH_DAY8_EFF_VS_DAY15_EFF_CD8_TCELL_UP"      
 [8] "KAECH_DAY8_EFF_VS_DAY15_EFF_CD8_TCELL_DN"      

This indicates that authorship, time (or other experimental design factors), cell type, and regulatory association may be encoded in standard gene set names. The authorship information can be retrieved systematically from the PMID attribute, and its role in the set name can probably be ignored.

Can we parse the set names, using the fact that underscore separates the key descriptive tokens? Perhaps, but examples like GSE19888_ADENOSINE_A3R_INH_PRETREAT_AND_ACT_ BY_A3R_VS_A3R_INH_AND_TCELL_ MEMBRANES_ACT_MAST_CELL_UP are not encouraging.

The DESCRIPTION_BRIEF field is more human-readable. Here are examples associated with the first four standard names listed above.

[1] "Genes up-regulated in naïve CD8 T cells compared to effector CD8 T cells at the peak expansion phase (day 8 after LCMV-Armstrong infection)."  
[2] "Genes down-regulated in naïve CD8 T cells compared to effector CD8 T cells at the peak expansion phase (day 8 after LCMV-Armstrong infection)."
[3] "Genes up-regulated in naïve CD8 T cells compared to effector CD8 T cells at contraction phase (day 15 after LCMV-Armstrong infection)."        
[4] "Genes down-regulated in naïve CD8 T cells compared to effector CD8 T cells at contraction phase (day 15 after LCMV-Armstrong infection)."   

Some readers may find it a little challenging to extract the key distinctions among these sets -- up and down regulation are distinguished, but so are phases (peak expansion vs contraction), cell types, and times after LCMV-Armstrong infection.

The purpose of this package is to mobilize text mining technologies to help users interrogate gene set information in MSigDb through verbalization of relevant scientific concepts.

Illustration with the C7 (immunologic) gene set collection

We have created a DocSet instance for 300 gene sets from the immunologic set collection.

immu300

This is essentially a collection of environments mapping from gene set names and descriptions to genes and back. We have a searchDocs method to query the DocSet.

dsl = searchDocs("down.*systemic lupus", immu300)
datatable(dsl)

It is possible to search for occurrence of a gene in the gene sets catalogued in the DocSet instance.

brocc = searchDocs("BRCA2", immu300)
brocc

TO DO

The abstracts (DESCRIPTION_FULL fields) of the sets need to be introduced. This would entail enhancement to the ssrch package.

Need to do the full c7 and see how large the resulting DocSet is.

Do other classes.

vjcitn/msigTextM documentation built on May 13, 2019, 12:33 a.m.