I just reread all of the last 25 posts you have submitted (going back to Oct 27th, 2011) and in those 25 posts there is not one line of code that you have shown us that you have written on your own.

It is true that your last post originally contained some code that you had written, but you edited your post and removed that code; so I can't count that. (I admit that I didn't look at the rest of your posts to try to determine if you had posted and removed other code.)

If you are unwilling to show us that you are making any attempt to write code for yourself and just ask us to write code for you; what about my statement:
was uncalled for? If you were in the habit of showing us your work and asking us to help fix parts that didn't work in your earlier posts, why did you stop? We are happy to help you learn why your code doesn't work. And, if you do try to write code on your own, you'll learn how to do it yourself MUCH faster.

Now, moving on to the problem at hand. With your sample input, there are 795 samples. So, to get more than 90%, you need 718 samples with a single non-dash value. For character position 47 (where there is a dash in the last input line), the counts are:

Code:

-  17
A 700
C   0
G   0
T  78

and, therefore, by your rules, the dash in that line should not be changed.

If you agree that the last line should not change, the following code seems to do what you want:

Code:

awk '
FNR == 1 && cnt {
	p90 = tot * .9
#	printf("p90=%f, tot=%d\n", p90, tot)
#	for(i in cc) printf("cc[%s]=%d\n", i, cc[i])
	for(i in cc) {
		if(cc[i] > p90) {
			off = index(i, SUBSEP)
			rep[substr(i, 1, off - 1)] = substr(i, off + 1)
		}
		delete cc[i]
	}
#	for(i in rep) printf("rep[%s]=%s\n", i, rep[i])
}
FNR == NR {
	if($2 == "Freq") {
		cnt = $3
		tot += cnt
	} else	for(i = length($0); i > 0; i--) {
			if((c = substr($0, i, 1)) == "-") continue
			cc[i, c] += cnt
		}
	next
}
NF == 1 {
	for(i = length($0); i > 0; i--)
		if((substr($0, i, 1) == "-") && (i in rep))
			$0 = (i > 1 ? substr($0, 1, i - 1) : "") rep[i] \
				substr($0, i + 1)
				
}
1' Input.txt Input.txt

producing the output:

Code:

>39sample Freq 4
TAGATGTGCCCGTGGGTTTCCCGTCAACACCGGATAGTAGCAGCACTA
>22sample Freq 15
TTGATGTCGTGGGTTTCCCGTCAACACCGGCAAATAGTAGCAGCACTA
>sample2 Freq 50
TCGATGTGCCAGTTTCCCGTCTAGCAGCACTACCAGGACCTTCGCCTA
>sample1 Freq 700
TTGATGTGTCCCGTCGACACCGGCAAATAGCAGCAGCACTACCAGGAC
>1-2sample Freq 9
TCGATTCCCGTCGACACCGGCAAATAGCAGCAGCACTACAGGACCTTC
>HL8OVD01AV2U9 Freq 17
TTGATGTGCCAGCTTTCGCGTCGACACCGGCAAATAGTAGCAGCGC-A

If you want to try this on a Solaris/SunOS system, change awk to /usr/xpg4/bin/awk.

With large input files (especially when there are very few dashes in the input), the last for loop in this script could be made much more efficient. For files with a few hundred input lines with line lengths around 50 characters, it might not be worth the time needed to improve and test it.

If you think it needs to run faster and think fixing this loop would help, feel free to look for dashes instead of evaluating each character position. If you try enhancing this script to do that and run into problems, we'll be happy to help you debug your code.