Replace:

    my $parent = $proteinLU{$word};
    my $wIndex = 0;

    while ($wIndex < @words && exists $parent->{$words[$wIndex]}) {
        $parent = $parent->{$words[$wIndex++]}
    }

    print "**$parent->{_name_}** " if exists $parent->{_name_};
    splice @words, 0, $wIndex;
[download]

with:

    my $parent = $proteinLU{$word};
    my $wIndex = 0;
    my @best;

    while ($wIndex < @words && exists $parent->{$words[$wIndex]}) {
        @best = ($parent->{_name_}, $wIndex) if exists $parent->{_name
+_};
        $parent = $parent->{$words[$wIndex++]};
    }

    if (@best) {
        print "**$best[0]** ";
        splice @words, 0, $best[1];
    } else {
        print "$word ";
    }
[download]

There was no fall back to the longest match so far if a longer partial match existed. In the sample case 'MCM2' was being masked by 'MCM2 , FORMERLY'.

There is no need (but no major harm) to provide different case versions of the same match string unless you want to use a case sensitive match with only the given variants allowed (in which case you need to remove lc in the various places it is used).