CODEWITHSUNDEEP
rmergefuzzyjoin
Beeba
Beeba

Reputation: 642

Merging two tables where one column is substring of the other in R

I have two data.frames with columns that contain accession numbers

subset of df 1:

sub_df1 <- structure(list(database = "CLO, ArrayExpress, ArrayExpress, ATCC, BCRJ, BioSample, CCLE, ChEMBL-Cells, ChEMBL-Targets, Cosmic, Cosmic, Cosmic, Cosmic-CLP, GDSC, GEO, GEO, GEO, IGRhCellID, LINCS_LDP, Wikidata", 
    database_accession = "CLO_0009006, E-MTAB-2770, E-MTAB-3610, CRL-7724, 0337, SAMN03471142, SH4_SKIN, CHEMBL3308177, CHEMBL2366309, 687440, 909713, 2159447, 909713, 909713, GSM887568, GSM888651, GSM1670420, SH4, LCL-1280, Q54953204"), .Names = c("database", 
"database_accession"), row.names = 2L, class = "data.frame")

subset of df 2:

sub_df2 <- structure(list(database_accession = "SH4_SKIN", G1 = -1.907138, 
    G2 = -7.617305, G3 = -3.750553, G4 = 2.615004, G5 = 9.751557), .Names = c("database_accession", 
"G1", "G2", "G3", "G4", "G5"), row.names = 101L, class = "data.frame")

I would like to merge the two dataframes by the column database_accession but the problem is they are not exact matches. the string insub_df2 is a substring of the string in sub_df1.

I thought about using fuzzyjoin but having a hard time getting the match algorithm right.

Upvotes: 4

Views: 2407

Answers (2)

Arthur Yip
Arthur Yip

Reputation: 6230

The fuzzyjoin solution, using match_fun = str_detect or regex_join():

library(tidyverse); library(fuzzyjoin)
# Load data
sub_df1 <- structure(list(database = "CLO, ArrayExpress, ArrayExpress, ATCC, BCRJ, BioSample, CCLE, ChEMBL-Cells, ChEMBL-Targets, Cosmic, Cosmic, Cosmic, Cosmic-CLP, GDSC, GEO, GEO, GEO, IGRhCellID, LINCS_LDP, Wikidata", database_accession = "CLO_0009006, E-MTAB-2770, E-MTAB-3610, CRL-7724, 0337, SAMN03471142, SH4_SKIN, CHEMBL3308177, CHEMBL2366309, 687440, 909713, 2159447, 909713, 909713, GSM887568, GSM888651, GSM1670420, SH4, LCL-1280, Q54953204"), .Names = c("database", "database_accession"), row.names = 2L, class = "data.frame")
sub_df2 <- structure(list(database_accession = "SH4_SKIN", G1 = -1.907138, G2 = -7.617305, G3 = -3.750553, G4 = 2.615004, G5 = 9.751557), .Names = c("database_accession", "G1", "G2", "G3", "G4", "G5"), row.names = 101L, class = "data.frame")

# Solution
# Using fuzzy_join. Could also use regex_full_join(), which is the wrapper for match_fun = str_detect, mode = "full"
fuzzy_join(sub_df1, sub_df2, match_fun = str_detect, by = "database_accession", mode = "full") %>% 
  str()
#> 'data.frame':    1 obs. of  8 variables:
#>  $ database            : chr "CLO, ArrayExpress, ArrayExpress, ATCC, BCRJ, BioSample, CCLE, ChEMBL-Cells, ChEMBL-Targets, Cosmic, Cosmic, Cos"| __truncated__
#>  $ database_accession.x: chr "CLO_0009006, E-MTAB-2770, E-MTAB-3610, CRL-7724, 0337, SAMN03471142, SH4_SKIN, CHEMBL3308177, CHEMBL2366309, 68"| __truncated__
#>  $ database_accession.y: chr "SH4_SKIN"
#>  $ G1                  : num -1.91
#>  $ G2                  : num -7.62
#>  $ G3                  : num -3.75
#>  $ G4                  : num 2.62
#>  $ G5                  : num 9.75

Created on 2019-03-17 by the reprex package (v0.2.1)

Upvotes: 4

IceCreamToucan
IceCreamToucan

Reputation: 28705

You can use the sqldf package and write a query joining the tables with a like condition to test if the value in sub_df1 contains the value in sub_df2.

library(sqldf)
sqldf('
select  *
from    sub_df2 two
        left join sub_df1 one
          on one.database_accession like "%" || two.database_accession || "%"
')

Upvotes: 1

Related Questions