UNIX for Dummies Questions & Answers

I am processing RNA-seq data files that have been aligned using RUM. One of the output files is a *.sam that includes:

Unique alignments
Non-unique alignments
original read files

I want to extract only the unique alignments by pulling out alignments that have "IH:i:1" (indicates this read had only one alignment).

I have tried.... grep "IH:i:1" file.sam > filtered.sam

but this will also return "IH:i:11 IH:i:12" etc etc. I have also tried "IH:i:1 " which returns nothing (I believe it's tab delimited)

Any suggestions?

Try with grep -w to match exact word
grep -w 'IH:i:1' file.sam | uniq -u >filtered.sam

As you didn't post a sample of your file i am left to guessing, but no problem: just give me a moment while i inspect your file with my crystal ball....

OK, somehow the device must be broken, but i can still give you some pointers:

"grep" always searches for the expression you give it. "1" will find "1", but also "11", "12", "1blabla", etc.. The trick therefore is to make to expression longer, up to a point where all the wrong lines are excluded.

Having said this: if you are sure (instead of just guessing) that the expression is followed by a tab you could add this tab as part of the expression. grep accepts this, but you will have to enclose your expression in single quotes to protect it from the shell evaluation process. In the following "<TAB>" means a literal tab character:

Code:

grep 'IH:i:1<TAB>' /path/to/infile > filtered.file

You can verify the character following being a tab by opening the file in a "vi" editor and enter ":set list" to display all the non-printable characters. Tabs will be displayed as "^I".

In case you just don't want any digit to follow your expression you simply forbid that instead of trying to find out what follows your expression:

Code:

grep 'IH:i:1[^0-9]' /path/to/infile > filtered.file

"[^0-9]" means "any character except numbers 0-9". The "[...]" means "any one character inside" and the "^" at the beginning reverses the meaning.

I hope this helps.

bakunin