| Analysis | We defined a read pair as a diagnostic paired-end (PE) if the two ends of a read pair 1) could both be aligned but 2) could not meet the pair-end insert size and/or orientation requirement. We grouped abnormally mapped paired-end reads with coordinate distances smaller than the maximum insert size on both ends into diagnostic PE clusters. In order to avoid misalignment, PE clusters with greater than 4 pairs were discarded. Common structural variations like deletions, translocations, duplications, inversions etc. were examined and summarized into alignment models. This paired-end method (PEM) is substantially biased to deletion events. In addition to searching for SVs using paired-end methods (PEM), we also identified copy number variations (CNV) based on read depth. By modeling sequencing depth distribution on different levels of GC content, we found 1701 CNV regions that had a lower copy number (1-47 kb in length, median at 1 kb) and 1299 that had a higher copy number (1-105 kb in length, median at 1 kb) than NCBI36. Approximately 82% of the CNV regions that had a lower copy number and 61% CNVs that had a higher copy number had more than a 50% overlap with the annotated repeats, which was not surprising. However, these regions may be somewhat inaccurate because sequencing depth can be affected by many factors, including GC content and alignment difficulties in repetitive region. |
