Extract rows and substrings from one file conditional on information of another file

问题

I have a file 1.blast with coordinate information like this

1       gnl|BL_ORD_ID|0 100.00  33      0       0       1        3
27620   gnl|BL_ORD_ID|0 95.65   46      2       0       1       46
35296   gnl|BL_ORD_ID|0 90.91   44      4       0       3       46
35973   gnl|BL_ORD_ID|0 100.00  45      0       0       1       45
41219   gnl|BL_ORD_ID|0 100.00  27      0       0       1       27
46914   gnl|BL_ORD_ID|0 100.00  45      0       0       1       45

and a file 1.fasta with sequence information like this

>1
TCGACTAGCTACGACTCGGACTGACGAGCTACGACTACGG
>2
GCATCTGGGCTACGGGATCAGCTAGGCGATGCGAC
...
>100000
TTTGCGAGCGCGAAGCGACGACGAGCAGCAGCGACTCTAGCTACTG

I am searching now a script that takes from 1.blast the first column and extracts those sequence IDs (=first column $1) plus sequence and then from the sequence itself all but those positions between $7 and $8 from the 1.fasta file, meaning from the first two matches the output would be

>1
ACTAGCTACGACTCGGACTGACGAGCTACGACTACGG
>27620
GTAGATAGAGATAGAGAGAGAGAGGGGGGAGA
...

(please notice that the first three entries from >1 are not in this sequence)

The IDs are consecutive, meaning I can extract the required information like this:

awk '{print 2*$1-1, 2*$1, $7, $8}' 1.blast

This gives me then a matrix that contains in the first column the right sequence identifier row, in the second column the right sequence row (= one after the ID row) and then the two coordinates that should be excluded. So basically a matrix that contains all required information which elements from 1.fasta shall be extracted

Unfortunately I do not have too much experience with scripting, hence I am now a bit lost, how to I feed the values e.g. in the suitable sed command? I can get specific rows like this:

sed -n 3,4p 1.fasta

and the string that I want to remove e.g. via

sed -n 5p 1.fasta | awk '{print substr($0,2,5)}'

But my problem is now, how can I pipe the information from the first awk call into the other commands so that they extract the right rows and remove from the sequence rows then the given coordinates. So, substr isn't the right command, I would need a command remstr(string,start,stop) that removes everything between these two positions from a given string, but I think that I could do in an own script. Especially the correct piping is a problem here for me.

回答1:

As either thunk and msw have pointed out, more suitable tools are available for this kind of task but here you have a script that can teach you something about how to handle it with awk:

Content of script.awk:

## Process first file from arguments.
FNR == NR {
        ## Save ID and the range of characters to remove from sequence.
        blast[ $1 ] = $(NF-1) " " $NF
        next
}

## Process second file. For each FASTA id...
$1 ~ /^>/ {
        ## Get number.
        id = substr( $1, 2 )

        ## Read next line (the sequence).
        getline sequence

        ## if the ID is one found in the other file, get ranges and
        ## extract those characters from sequence.
        if ( id in blast ) {
                split( blast[id], ranges )
                sequence = substr( sequence, 1, ranges[1] - 1 ) substr( sequence, ranges[2] + 1 )
                ## Print both lines with the shortened sequence.
                printf "%s\n%s\n", $0, sequence
        }

}

Assuming your 1.blasta of the question and a customized 1.fasta to test it:

>1
TCGACTAGCTACGACTCGGACTGACGAGCTACGACTACGG
>2
GCATCTGGGCTACGGGATCAGCTAGGCGATGCGAC
>27620
TTTGCGAGCGCGAAGCGACGACGAGCAGCAGCGACTCTAGCTACTGTTTGCGA

Run the script like:

awk -f script.awk 1.blast 1.fasta

That yields:

>1
ACTAGCTACGACTCGGACTGACGAGCTACGACTACGG
>27620
TTTGCGA

Of course I'm assumming some things, the most important that fasta sequences are not longer than one line.

回答2:

If you do bioinformatics and work with DNA sequences (or even more complicated things like sequence annotations), I would recommend having a look at Bioperl. This obviously requires knowledge of Perl, but has quite a lot of functionality.

In your case you would want to generate Bio::Seq objects from your fasta-file using the Bio::SeqIO module.

Then, you would need to read the fasta-entry-numbers and positions wanted into a hash. With the fasta-name as the key and the value being an array of two values for each subsequence you want to extract. If there can be more than one such subsequence per fasta-entry, you would have to create an array of arrays as the value entry for each key.

With this data structure, you could then go ahead and extract the sequences using the subseq method from Bio::Seq.

I hope this is a way to go for you, although I'm sure that this is also feasible with pure bash.

回答3:

This isn't an answer, it is an attempt to clarify your problem; please let me know if I have gotten the nature of your task correct.

foreach row in blast:
    get the proper (blast[$1]) sequence from fasta
    drop bases (blast[$7..$8]) from sequence
    print blast[$1], shortened_sequence

If I've got your task correct, you are being hobbled by your programming language (bash) and the peculiar format of your data (a record split across rows). Perl or Python would be far more suitable to the task; indeed Perl was written in part because multiple file access in awk of the time was really difficult if not impossible.

You've come pretty far with the tools you know, but it looks like you are hitting the limits of their convenient expressibility.

回答4:

Updated the answer:

awk  '
NR==FNR && NF { 
    id=substr($1,2)
    getline seq
    a[id]=seq
    next 
} 
($1 in a) && NF { 
    x=substr(a[$1],$7,$8)
    sub(x, "", a[$1])
    print ">"$1"\n"a[$1]
} ' 1.fasta 1.blast

来源：https://stackoverflow.com/questions/16836300/extract-rows-and-substrings-from-one-file-conditional-on-information-of-another

标签

string

bash

sed

awk

fasta