As BrowserUk pointed out here, there is no  ! regex metacharacter/operator (presumably meaning "not-followed-by-the-following-pattern"). What is probably meant is  (?!pattern) (see Look-Around Assertions in perlre; also see perlretut).

This may be what TamaDP was looking for (there are some other good solutions):

c:\@Work\Perl\monks>perl -wMstrict -le
"my @proteins = qw(
   DAAAAATTLTTTAMTTTTTTCKMMFRPPPPPGGGGGGGGGGGG
   ALTAMCMNVWEITYHKGSDVNRRASFAQPPPQPPPPLLAIKPASDASD
   AAAKAAA  AAAKAAA  XXXXXX
   );
 ;;
 my @new_peptides;
 for my $protein (@proteins) {
   if ($protein =~ s{ (?<= [KR]) (?! P) }{=}xmsg) {
     push @new_peptides, split ('=', $protein);
     }
   }
 ;;
 for my $peptide (@new_peptides) {
   print qq{Peptide is '$peptide'};
   }
"
Peptide is 'DAAAAATTLTTTAMTTTTTTCK'
Peptide is 'MMFRPPPPPGGGGGGGGGGGG'
Peptide is 'ALTAMCMNVWEITYHK'
Peptide is 'GSDVNR'
Peptide is 'R'
Peptide is 'ASFAQPPPQPPPPLLAIKPASDASD'
Peptide is 'AAAK'
Peptide is 'AAA'
Peptide is 'AAAK'
Peptide is 'AAA'
[download]

Yet another solution might be:

c:\@Work\Perl\monks>perl -wMstrict -le
"my @proteins = qw(
   DAAAAATTLTTTAMTTTTTTCKMMFRPPPPPGGGGGGGGGGGG
   ALTAMCMNVWEITYHKGSDVNRRASFAQPPPQPPPPLLAIKPASDASD
   AAAKAAA  AAAKAAA  XXXXXX
   );
 ;;
 my $cleave = qr{ (?<= [KR]) (?! P) }xms;
 ;;
 my @peptides =
   map  split($cleave),
   grep m{ $cleave }xms,
   @proteins
   ;
 ;;
 print qq{Peptide is '$_'} for @peptides;
"
Peptide is 'DAAAAATTLTTTAMTTTTTTCK'
Peptide is 'MMFRPPPPPGGGGGGGGGGGG'
Peptide is 'ALTAMCMNVWEITYHK'
Peptide is 'GSDVNR'
Peptide is 'R'
Peptide is 'ASFAQPPPQPPPPLLAIKPASDASD'
Peptide is 'AAAK'
Peptide is 'AAA'
Peptide is 'AAAK'
Peptide is 'AAA'
[download]

my @peptides =           # 4. and the pieces are peptides.
  map  split($cleave),   # 3. split at each cleavage point...
  grep m{ $cleave }xms,  # 2. that can be cleaved, ...
  @proteins              # 1. for each protein...
  ;
[download]

Update: After sober consideration, I changed the code example above to its present form. The previous code contained
    my $cleave = qr{ [KR] (?! P) }xms;
    ...
      map  split(m{ (?<= $cleave) }xms),
      grep m{ $cleave }xms,
      ...
which doesn't really respect the DRY principle, which is what I was aiming to exemplify.