MyDGR is built upon an improved version of DGRscan we previously developed (15 (link)). As shown in Figure 1A, a minimal DGR system consists of a RT gene and a TR-VR pair, and DGRscan was devised based on finding these core components. In particular, MyDGR uses the de novo search function in DGRscan (see Figure 1B). Given an input nucleotide sequence, MyDGR first identifies putative RT genes by searching the translated nucleotide sequence against a protein database of 155 RT proteins (21 ) (using blastx). If it finds putative RT genes, it then scans in the neighborhood of each of these putative RT genes (10 kb in both ends), searching for segments that potentially form a TR-VR pair: two repeats that are similar to each other spanning at least 60 bp with seven or more substitutions involving adenines in one of the repeats (i.e. the TR), allowing only a small fraction (≤30%) of the substitutions to be involved in non-As in the putative TR. Although rare, TR and VR may be on opposite strands in some genomes (3 (link),19 (link))—myDGR does not limit its search for TR–VR pairs on the same strand. A dynamic programming algorithm is used for aligning the candidate TR-VR pairs; however, to speed up the alignment process, a full dynamic programming is called only when a seed match of at least 60 bp (without indels) is found between two candidate segments. We also note that using putative RT as the constraint not only significantly reduces the search space of TR-VR pairs, but also helps eliminate potential false DGRs.