How to build data matrix from mixed and messy CSV file?

Question

I have a huge .csv file like this :

Transcript Id   Gene Id(name)   Mirna Name  miTG score
ENST00000286800 ENSG00000156273 (BACH1) hsa-let-7a-5p   1         


        

Answer 1

Using this test data:

Lines <- " Transcript Id   Gene Id(name)   Mirna Name  miTG score
ENST00000286800 ENSG00000156273 (BACH1) hsa-let-7a-5p   1
UTR3    21:30717114-30717142    0.05994568  
UTR3    21:30717414-30717442    0.13591267  
ENST00000345080 ENSG00000187772 (LIN28B)    hsa-let-7a-5p   1
UTR3    6:105526681-105526709   0.133514751"

read it all in and set the names, nms for the output. Then calculate the grouping vector, cs, using a cumulative sum. non-duplicates are the first row of each group and duplicates are the following rows. Merge these two sets of rows by group and extract out the highest MRE_score in each group:

DF <- read.table(text = Lines, header = TRUE, fill = TRUE, as.is = TRUE, 
         check.names = FALSE)
nms <- c("cs", names(DF)[1:5], "UTR3", "MRE_score") # out will have these names
DF$cs <- cumsum(!is.na(DF$Mirna)) # groups each ENST row with its UTR3 rows
dup <- duplicated(DF$cs) # FALSE for ENST rows and TRUE for UTR3 rows
both <- merge(DF[!dup, ], DF[dup, ], by = "cs")[c(1:6, 11:12)]  # merge ENST & UTR3 rows
names(both) <- nms
both$MRE_score <- as.numeric(both$MRE_score)
Rank <- function(x) rank(x, ties.method = "first")
out <- both[ave(-both$MRE_score, both$cs, FUN = Rank) == 1, -1] # only keep largest score

Here we get:

> out
       Transcript              Id     Gene      Id(name) Mirna                  UTR3 MRE_score
2 ENST00000286800 ENSG00000156273  (BACH1) hsa-let-7a-5p     1  21:30717414-30717442 0.1359127
3 ENST00000345080 ENSG00000187772 (LIN28B) hsa-let-7a-5p     1 6:105526681-105526709 0.1335148

Note that the question refers to a CDS column but what it is is not described nor does it appear in the example output so we ignored it.