use commandline arguments in a sed command

I'm not sure that $1, $2… and so on can be accessed from command line.

Instead, you could try to run your line from a function. You definitely can pass argument to functions.

foo() { sed 's/^\(.\{'"$2"'\}\)./\1'"$3"'/' "$1"; }foo out2.fa 2 A

Edit: As suggested in comments, I added double-quotes around the argument references to prevent spaces from breaking command parsing (That should not be necessary around $2 as we just pass an integer, but it's much better for $1 and $3.

Update:

While what I say after the horizontal line (the original answer) is true in general, the real problem is the OP's mistaken belief that positional string arguments ($1, ...) can be passed to sed with arguments after the 1st post-script (filename) argument.
All arguments after sed's script argument are, in fact, filename arguments, and sed has no concept of other types of arguments or shell variables in general - any references to shell variables or arguments must be expanded BEFORE the script is passed to sed.
Generally, arguments $1, ... are shell-only constructs that refer to the arguments passed to either the current shell or, inside a shell function, that function.

• @Qeole's answer provides a solution based on a shell function, which is the best option.
• Alternatively, use variables:

countBefore=2 newChar='A'sed 's/^\(.\{'"$countBefore"'\}\)./\1'"$newChar"'/' out2.fa

Note:

• You say changes the 2nd position to an A. Actually, your command effectively replaces the 3rd character ($countBefore+1-nth).)
• You say, I am currently wrapping the arguments in ''. Actually, it's the sed script that's enclosed in '', and you're splicing in unquoted argument references. My variable-based solution double-quotes the spliced-in variable references for robustness.

• The alternative to splicing in variable references is to simply double-quote the entire sed script and use embedded variable references:

  sed "s/^\(.\{$countBefore\}\)./\1$newChar/" out2.fa

sed: -e expression #1, char 21: Invalid content of \{\} is GNU sed's way of saying that what you specified to go inside {...} ($1) does not amount to a valid repetition-count expression.

In other words: the value of $1 in your case is NONE of the following:

• a mere decimal number (e.g., 2)
• a decimal number followed by ,, optionally followed by another decimal number (e.g., 2, or 2,3)
• ,, followed by a decimal number (e.g., ,3)

Make sure that $1 is one of the above.