I am trying to find the number of reads that align uniquely, to two, to three and multilocations in a bam file.

This bam was create using PE reads and below is two lines from the SAM file:

HWI-D00123R2:155:C4AM3ACXX:2:1103:4696:31199    403 AL2G13910.t1    5619    60  100M    =   5586    -133    GAGATAGTGGAATCATCCGAACCCTGGATGTCCCGATCTATATCACCAAGGTGTCTGGTAATACAATCTTCTGCTTGGATCGGGATGGGAAAAACAAGGC    DDDDCDDEDDEEDDDDDDDAADDDEEEEDFFHHHJJIJJJIIGF=JIJIHJJIIIJJJJJJIIJJJJJJJJJIJJJJJJJJJJJJJJHHHHHFFFFFCCC    NH:i:1  HI:i:1
HWI-D00123R2:155:C4AM3ACXX:2:1103:4698:74520    419 AL2G30660.t1    2065    60  88M =   2141    176 CTGAACCTGCCATCGCTGGAAACGTATCTGCTGCCTCACCAGTTGATGACAAGAACGATGATGGAGATGAACATCACGAGATCGATCT    GJJJIJJJJJJJJJJJJJJJJJJJHIJJJJJJJJJJJIHHIJJJJJJJJJJJJJJHHHFFFFFEEEEEDDDDDDDDDDDDDDDDDDDD    NH:i:1  HI:i:1

To count unique alignments I used something like:

grep NH:i transcripts_test.sam | wc -l

Well, NH:i:... is the tag suggest to how many locations was the read aligned to. But, since the bam is based on PE reads should that count be divided by two? If so the counts won't be right when only one mate is uniquely mapped, rite?.

And, is there a tool that I can use to prepare me alist of number of reads for each level of alignments?

Thanks,