如何根据组内的子串匹配两个数据帧

Question

我想合并两个数据帧，这两个数据帧由相同的标识符分组。第一个数据帧(valueA)中的变量应与第二个数据帧(valueB)中的变量的子字符串匹配，但仅限于组内。

我可以设法匹配匹配的变量，但我正在努力将匹配限制到分组变量。以下是样本数据帧和匹配的代码：

df1 <- data.frame(report = c('Report1','Report1','Report1','Report1','Report1','Report1'),
            identifier = c('Abraham', 'Abraham', 'Abraham','Barack', 'Barack','Barack'),
            variableA = c('V1','V2','V3','V1','V2', 'V3'),
            value = c('CDKN2A/B','PALB2','KRAS','TP53','RB1','KRAS'))

df2 <- data.frame(report = c('Report1','Report1','Report1','Report1','Report1','Report1','Report1'),
            identifier = c('Abraham', 'Abraham', 'Abraham','Abraham','Barack', 'Barack','Barack'),
            variableB = c('F1','F2','F3','F4','F1','F2', 'F3'),
            valueB = c('CDKN2A/B LOSS','PALB2 P1111FS*13','KRAS G12R','PALB2 N540FS*1','RB1 SPLICE SITE 2325+1G>A','KRAS G13C','TP53 C238F'))

下面是我尝试过的代码，但它不适用于组

idx2 <- sapply(df1$value, grep, df2$valueB)
idx1 <- sapply(seq_along(idx2), function(i) rep(i, length(idx2[[i]])))
idx3 <- cbind(df1[unlist(idx1),,drop=F], df2[unlist(idx2),,drop=F])

预期输出为(数据帧的代码)

df3 <- data.frame(report=c('Report1','Report1','Report1','Report1','Report1','Report1','Report1'),
                  identifier = c('Abraham', 'Abraham', 'Abraham','Abraham','Barack', 'Barack','Barack'),
                  variableA = c('V1','V2','V3','V2','V1','V2', 'V3'),
                  value = c('CDKN2A/B','PALB2','KRAS','PALB2','TP53','RB1','KRAS'),
                  variableB = c('F1','F2','F3','F4','F1','F2', 'F3'),
                  valueB = c('CDKN2A/B LOSS','PALB2 P1111FS*13','KRAS G12R','PALB2 N540FS*1','TP53 C238F','RB1 SPLICE SITE 2325+1G>A','KRAS G13C'))

结果数据帧

report  identifier  variableA   value   variableB   valueB
Report1 Abraham     V1      CDKN2A/B    F1  CDKN2A/B LOSS
Report1 Abraham     V2      PALB2   F2  PALB2   P1111FS*13
Report1 Abraham     V3      KRAS    F3  KRAS    G12R
Report1 Abraham     V2      PALB2   F4  PALB2   N540FS*1
Report1 Barack      V1      TP53    F1  TP53    C238F
Report1 Barack      V2      RB1 F2  RB1 SPLICE SITE 2325+1G>A
Report1 Barack      V3      KRAS    F3  KRAS    G13C

希望这是有意义的，非常感谢你的帮助！

Answer 1

您可以使用fuzzyjoin-package执行以下操作：

fuzzy_inner_join(df2, df1, by = c("valueB" = "valueA", "identifier" = "identifier"), match_fun = list(str_detect, `==`)) %>%
  select(report.x, identifier.x, variableA, valueA, variableB, valueB)

  report.x identifier.x variableA   valueA variableB                    valueB
1  Report1      Abraham        V1 CDKN2A/B        F1             CDKN2A/B LOSS
2  Report1      Abraham        V2    PALB2        F2          PALB2 P1111FS*13
3  Report1      Abraham        V3     KRAS        F3                 KRAS G12R
4  Report1      Abraham        V2    PALB2        F4            PALB2 N540FS*1
5  Report1       Barack        V2      RB1        F1 RB1 SPLICE SITE 2325+1G>A
6  Report1       Barack        V3     KRAS        F2                 KRAS G13C
7  Report1       Barack        V1     TP53        F3                TP53 C238F

这样，您就可以对不同的列应用不同的匹配函数。在本例中，我们对模糊匹配列使用str_detect()，对分组列使用==。

Answer 2

我们可以使用str_extract来捕获公共字符串并在其上进行合并，即

library(stringr)

merge(df1, 
      transform(df2, value = sapply(df2$valueB, function(i)
                                  str_extract(i, paste(df1$value, collapse = '|')))), 
       by = c('value', 'identifier', 'report'))

#         value identifier  report variableA variableB                    valueB
#    1 CDKN2A/B    Abraham Report1        V1        F1             CDKN2A/B LOSS
#    2     KRAS    Abraham Report1        V3        F3                 KRAS G12R
#    3     KRAS     Barack Report1        V3        F2                 KRAS G13C
#    4    PALB2    Abraham Report1        V2        F2          PALB2 P1111FS*13
#    5    PALB2    Abraham Report1        V2        F4            PALB2 N540FS*1
#    6      RB1     Barack Report1        V2        F1 RB1 SPLICE SITE 2325+1G>A
#    7     TP53     Barack Report1        V1        F3                TP53 C238F