Rather than splitting on whitespace since there is whitespace in the descriptions as well, why not just split the whole data chunk with a regex from the start. Assuming that your entries are always formatted with the chr, start, end, and description, you could do something like:

my ( $chr, $start, $end, $description ) = $annotation_line =~ /(chr\d+
+) (\d+) (\d+) (.*)/;
if ( $description =~ /dgv_Cnv=True,overlap=\d+\.\d+%;/ ) {
    print "$chr\t$start\t$end\t$&\n";
}
[download]