Gravar-mail: Does replication groups scoring reduce false positive rate in SNP interaction discovery?